The work list

Left by the completed walk of the 1836–1869 death register — see row 69. 1. ~~«Franz 1853» may be Franziska.~~ CLOSED 27 September 2026 — she is Franziska. The 1832–1870 baptismal register was re-opened at book p. 130: 19 September 1853 · Obr. Feichting · house 9 · FRANCISCA · ehelich, father Johann Kozina, Hüblar, mother Ursula Sedej — and the stroke stands in the MÄDCHEN column. One child, not two. The archive's «Franz 1853» was a misreading of a short entry in the index; the death register, which had her as *Franziska, weiblich*, dead of measles at eleven in that same house, had been right all along. 2. ~~The five-Kozina count in that baptismal volume is short.~~ CLOSED the same day, and it cost a new person. The whole K section of the index (sig. 04632, entries 1–369, 1832–1870) was read end to end. The volume holds six Kozina, not five: Anton 1834 p.18 · Petrus 1836 p.33 · Martin 1841 p.64 · Francisca 1853 p.130 · Ana 1856 p.144 · Maria 1859 p.159. Maria 1859 was the missing one — born and baptised 7 August 1859 at Oberfeichting 9 to Johann Kozina and Ursula Sedej, a sister this archive did not have. It was not Joseph. The boy who died at two at that house in 1843 is not in this index at all, so he was baptised elsewhere or the index missed him, and neither is proved. 3. ~~Ursula Kosina's date is partial.~~ CLOSED 27 September 2026 — she died 28 March 1845. Book p. 43 was re-opened and read from its own head down: the printed year at the top of the column is 1845, and the month headings run Jänner (29th), Februar (26th), März (2nd) — after which the days run 2, 4, 6, 12, 13, 16, 18, 18, 25, 28, 30 with no further heading, the last of them buried «1. April». Hers is the 28th, inside March; buried the 30th. And the page gave something the row did not have. The word under her name is ABNEHMERIN — the female of *Abnehmer*, the same institution this register calls Franz an *Ausnehmer* for in 1849 and Maria Dortschen an *Auszüglerin* for in 1841: a householder who has handed the house over and lives on the reserved portion. She had held Oberfeichting 13 and given it up. The third Kosina house in that village was hers. All three are now closed. And the lead ran too. Marianna Kosina, 10 January 1840, Breg 12, aged 26 and unmarried, was thought to be perhaps the «Kosina Marianna — 16» the 1812–31 baptismal index lists under 1813 — an entry this archive had reported as not on page 16. It is on page 16. Re-read at nine times magnification: **«12 August 1813 · Haus 27 · MARIANNA · ehelich · Vater: LUKAS KOSINER, *Kaischler* · Mutter: LUZIA O[k]lin[t]sh»**. The age fits to the month and *Keuschler* is the word she dies under; she is almost certainly the same woman, and she is Lukas Kosina's daughter — a sister of the 1818 Gertraud this archive had to rule out before Martin's mother could be settled. Not proved: the 1813 entry gives a house number and no village. The correction it cost: page 16 had been read at 2.4×, and that was not enough for the hand. Page 8 is still open — two Dorotheas, to Primus Wit[nitz] and to Johann Holz, neither of this name.

Asked on 23 September 2026: put the photographs of Tereza Žubrinić on the pages that talk about her. Both were already in the archive — `4500806-tereza-zubrinic.jpg`, a studio portrait, and `4502256-tereza-grave-senj.jpg`, her stone in Senj — and both were already on her own page and in the gallery. Neither was on the Kosina page, which devotes a whole section to that grave. Both are on it now. AND READING THE SECOND ONE PROPERLY CORRECTED THE PAGE. Standing against the tomb is a wooden cross with a black plaque: «OVDJE POČIVA U MIRU BOŽJEM · vlč. ANTUN KOSINA · ✱ 1921 † 2003». *POČIVA* is singular — here rests — not the *POČIVAJU* of the stone behind it, and not the *IN MEMORIAM* his sisters' plaque gives him. This page said twice that Antun Kosina does not lie in Senj. He does. He restored his grandmother's grave in 1991 and was buried in it in 2003. His name was never cut into the granite; the wooden marker is all there is, and it was still standing twenty-one years later when the photograph was taken. Confirmed by the account holder the same day: «anton is in that grave, yes» — so the stone and the family agree. THE UNCOMFORTABLE PART. The archive has held that photograph since October 2024 and read the stone out of it in September 2026 — transcribing Tereza, Zorica Kasumović above her, and a *second* wooden cross at the foot of the same tomb for Branko Blažević — and stepped over this one, which is in the same frame. A photograph is not read until everything in it is. What this says about the other 101 images. The media audit of 23 September found every one of them reaches a page, which is the shallow question. The deep one is whether anybody has read what is IN them, and this row is one worked example of the answer being no. Row 64 should be re-scoped from «are they cited» to «are they read».

Asked and answered on 23 September 2026. A sibling session measured that this archive draws 2,497 children and links 2,167 — 87% — against five archives at 100%, and put it as the estate's last content gap. Checking it changed the size of it. The 330 are *entries*, not people, and they are three different things: * 64 living people this archive withholds — no name, no page, by policy. They can never become links. And the 100% comparison is not like-for-like: Falco publishes a name-only page for each of its 106 living people, so every child links. This archive scores lower for showing the gap instead of filling it. * 27 MyHeritage sorting buckets — «AUSTRIA: De Franceschi», «ZAGREB Defranceski's». Labels, not people, and already gated against publication. * 155 real people, not one of whom carries one of the eight surnames: Defranceschi 37, Pavelić 32, Franceschi 9, Miletić 9, Grgurić 6, Grandić 5, Krpan 5, Untersinger 4. The account holder chose: publish none of them here, and link the Defranceschi ones across. THAT HALF IS BLOCKED, AND NOT BY THIS ARCHIVE. The Defranceski archive publishes nine people — Carlo, Camillo and the other documented figures of the name — and none of the 46 is among them. Checked by name against its `people.json`: zero matches. There is nothing to link to, and an empty link would have looked like progress. So the other half was done instead: /people/ now decomposes the 330 and says why each part is not a link, computed rather than typed. WHAT WOULD UNBLOCK IT. The Defranceski archive publishing person pages for the Defranceschi of this tree — which is that archive's decision and not this one's. If it ever does, the 46 become cross-archive links and this row closes.

All 174 pages of the 1836–1869 death register read, 27 September 2026, after its letter-tabbed index had been proved incomplete. It holds nine Kozina deaths. The index lists two. MARIA «MINA» DORTSCHEN, p.27 — 11 April 1841, Oberfeichting 9, 60, *Auszüglerin*, the female of the *Ausnehmer* this same register calls Franz in the same house eight years later. She died seven months before her grandson Martin was born in that house. JOHANN KOZINA, p.167 — 10 October 1868, Zgornje Bitno 9, 64, still a *Häubler*. AND THE TWO SISTERS WHO NEVER LEFT. Maria, baptised 6 December 1818, died 6 January 1868 aged 49, *ledige Inwohnerin*. Gertraud died in the same house in 1880, *neomožena gostja*. Both unmarried, both entered as lodgers in the house they were born in, twelve years apart — and one of them had borne Martin there. Seven of the nine died at number 9, and Gertraud in the next volume makes eight. WHAT THIS LEAVES, as row 72: Franziska Kozina 1864 may be the child this archive records as *Franz 1853* read as a man; the baptismal volume's five-Kozina count is now short by at least two; Ursula's month and year are untranscribed; and Marianna 1840 at Breg 12 is unplaced.

A sweep on 22 September 2026, prompted by a fair question: «are your site pages keeping up with the research?» Mostly yes — and in two places, no. FOUND AND FIXED. * The Kosina stat cards said «10 people · 1841 earliest birth» and the dek said «four generations», under a table whose own first six rows include a man born in 1775. They were computed from the GEDCOM alone. Now 16 · 1775 · six, with each Carniolan row carrying an approximate birth year in `tools/kosina.py` so the earliest is derived and not typed. * `/impossible/` said «all 103 people here bearing a form of Ivan are men» and «25 of the 27 bearing a form of Lucija are women». True when written; the real figures are 124 and 50 of 54 since publication widened. Both now read `names.json` instead of being typed. * `/senj-1943/` still said Otilija's death had «no day recorded» — it is 11 November 1943 — and gave Oskar only «1954». Both corrected, with Antun's mission and Neda's «žrtva zračnog napada» added. CHECKED AND SOUND. Every other stat card on the site is computed: `[family]`, `/pictures/`, `/searched/`, `/sources/`. `/senj-1943/`'s three are historical facts about the raid and correctly fixed. `/found/`, `/searched/` and `/coverage/` are data-driven and current. `connected.astro` already carried Kranj. FLAGGED AND NOT TOUCHED, because the definition behind each number could not be reproduced and guessing at it is how a correction becomes an error. Publication grew 1,460 → 1,903 on 22 September, so any prose counting *published* people may have moved: | page | claim | |---|---| | `/zubrinic-roots/` | «218 published people hang off one» — 400 Žubrinić are published now | | `/gazetteer/` and `/harvest/` | «house 22 holds 67 published people between 1840 and 1890» | | `/name/` | «the 79 people this tree writes *Papa*» — 120 Papić-ish are published now | | `/frontier/` | «53 people in this tree are born or die here» | The fix is the same in every case and it is not to retype the number: make the page ask the data. Each of these is one derivation away from being computed, and once computed it cannot drift again. --- 27 September 2026 — the four flagged claims are all computed now, and were checked in the built pages rather than in the source. `/zubrinic-roots/` renders «37 sorting buckets … 286 published people hang off one … 144 have no other parent»; `/gazetteer/` renders «house 22 holds 67 published people between 1840 and 1890», with the retracted 43 kept beside it; `/name/` renders «the 75 people this tree writes Papić (Papa)»; `/frontier/` renders «123 people in this tree are born or die here». Each reads its data file and cannot drift again. The seventeen numbers still typed on purpose are declared in `sources/typed-numbers.psv` and held by `check:numbers`, which refuses any undeclared hand-typed count on a page that imports data. They are facts about the world, results from somebody else's index, quotations of what this archive used to say, and arithmetic that cannot move. This row is closed.

SENT 21 September 2026 to `kustodijazadar@gmail.com` — Franjevačka kustodija sv. Jeronima, Trg sv. Frane 1, Zadar; kustos fra Tomislav Šanko. Why Zadar and not Buenos Aires, which is what was asked for. Croatian pastoral work in Argentina was this custody's from 1929, and fra Leonard Rusković, the first Croatian priest in Argentina, served from 1929 until his death in 1958 — so he was the priest on the ground in January 1954. The Hrvatsko vjersko središte sv. Nikole Tavelića did not yet exist in 1954, so Oskar's funeral is not in its books; it is in whatever the custody's own men kept. And the Buenos Aires end has just closed. In July 2026 the last Croatian friar in Argentina, fra Josip Peranić, went home after forty years. The community's own word for it was «Kako dalje?» Zadar is the address that will still be there next year, and the custody has just published a spomenica on the whole Argentine mission — «Tejar – Buenos Aires» — so somebody there has recently had the entire archive open on a desk. The letter asks for: any record of the funeral or death of January 1954; the name of Oskar's wife, which is what this family has wanted for decades; where he is buried; anything on the son born in Argentina who later went to Germany; and whether Franjo Kosina, the agronomist off the same ship, is a name they know. The centre itself: Av. dr. Ricardo Balbín 4925, 1430 Buenos Aires, +54 11 4542 3400. No email is verified for it — a city listing gives `tehar@ciudad.com.ar`, an old ISP domain, unconfirmed, and it was NOT written to. An unverified address is a stranger, and this letter carries family detail.

SENT 21 September 2026 to `base@cemla.com`, the database desk of CEMLA, Independencia 20, Buenos Aires. CEMLA's free search is now worked out: four spellings, forty-one rows, and exactly one is this family. «KOSINA · OSCAR · 25 · C · YUGOSLAVA/S.H.S. · SEGKA · COLONO · 1947/10/09 · PHILIPPA · GENOVA». The index gives one line. The bound book gives the page, and the page carries the entry number, the declared destination and whoever was receiving him. The letter asks three things. The full page for 9 October 1947. Whether any woman on that page or that sailing shares his declared destination or family-group number — he landed CASADO and no Kosina woman is in the database at all, so his wife is under a maiden name or she did not come. And the page of 1 April 1947 for Franjo Kosina, 47, an agronomist, who crossed on the same ship six months earlier: consecutive entry numbers or a shared sponsor would turn «possible brother» into evidence. The letter offers to pay their fee. Two doors remain and neither needs an archive. The reverse of Oskar's memorial card, which has never been scanned and on a card of that date almost certainly reads «Ožalošćena obitelj: supruga … sin …». And the Croatian chaplaincy in Buenos Aires, where fra Leonard Rusković served from 1929 until his death in 1958 and therefore buried him in January 1954.

SENT 21 September 2026 to `info@hkm-lu.de` — the Hrvatska katolička misija bl. Alojzija Stepinca at Ludwigshafen, the mission Ante Kosina founded in 1967 and led until 1985. It asks, in this order: 1. A photograph of him the family may publish. This archive has exactly one — the studio portrait in a cassock — and the mission will have more. 2. His ordination year and place. The family says the Gregoriana in Rome; no year is in hand. 3. His death in 2003 — the stone gives the year only. A date of 3 July 2003 is circulating from a generative summary and is in no source; the mission or the archdiocese can confirm or kill it. 4. Whether the mission holds anything of his — the 1967 founding papers, parish bulletins, letters. It offers in return everything this family has that they will not: his parents, his birth parish, the raid that killed his father and two sisters, and the grave he restored in Senj in 1991. The second door is the Riječka nadbiskupija, which ordained him and would hold his personnel file and an obituary. It has not been written to about him. The third is the Hrvatski dušobrižnički ured in Frankfurt, which coordinates the German missions.

All fifteen read, 15 September 2026, and question 4 4d is answered: the Michaels are at least seven men, in seven houses, in four villages. Šumećica N° 5 (four children, 1849–58), Šumećica N° 10, Poljica N° 13, Prozor N° 50, Obilje N° 7 and Prozor Br. 1 — six distinct wives. The *Conditio* column separates them and nothing else does. Šumećica N° 5 is a household this archive did not have — Michael Žubrinić and Maria Marković, four children across nine years. The Šumećica houses known before were N° 2, N° 10 and N° 24. Three Žubrinić women married Žubrinić men, and the index carries a given name only for every one of them. The index is also wrong at least once: it gives the 1852 mother at Šumećica N° 5 as «Maria Maković» where the register says Marković — a household that looked like two is one. The last entry, Maria of 28 November 1858, came in a third window: Prozor N° 49, next door to N° 50. Everything is in `sources/familysearch/otocac-michael-1846-1858.json`, and the cautions that cost this sitting time are in [notes/otocac-michael-house-numbers-2026-09-15.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/otocac-michael-house-numbers-2026-09-15.md) — the book changes language from page to page, the status word drifts while the house number holds, and entries hide in the last row of a left page below where a whole-page crop stops. Reconciled against the tree. The households are not new people: all six wives and nine of the fourteen children are already there. What the tree lacked was the house numbers — of those nine children exactly one carried one, *Nikola Miko Žubrinić*, «Prozor 50», which this reading confirms; the rest said «Otočac, Croatia» or nothing. Seven are now in `sources/readings.psv`, and five children genuinely absent from the tree are in `sources/found.psv`, every one of them with siblings the tree already has. A caution worth more than the finding. A first check concluded that none of the six wives was in the tree. That was wrong: they appear only as parent references on their children, with no page of their own, so a search of published people misses them. Searching `people.json` is not searching the tree, and a «not in the tree» claim made that way is worthless.

Found 15 September 2026, where this row expected a null. Ahnentafel 54 married ahnentafel 55 at Karlobag on 17 June 1807, folio 4, image 231: «…habito mutuo consensu *Georgii filij defti Georgii Uroda* et *Mariæ filiæ Michaelis Gergas* hujatium… præsentibus testibus *Caspare Jerko viduo* et *Joanne Fero…*». The date column reads 17 and the month is a ditto for the entry above it, *16 Junii*; the page runs 7 Jan · 14, 22, 26 *ejusdem* · 3 Feb · 7 May · 16 Jun · 17 · 8 Jul · 20 Sep, so the ditto can only be June. It proves the two brothers. The 1806 marriage of *Joannis filii defuncti Dñi Georgij Uroda* was read at the time as probably a brother rather than ahnentafel 54 himself, because ahnentafel 55's 1834 death entry calls her husband a *Georgius*. Confirmed from the other side: Joannes and Georgius are both sons of the same dead Georgius Uroda, married sixteen months apart in the same church. Ahnentafel 108 is named here as «defti Georgii Uroda», tying the 1791 burial to both marriages. Two cautions. The volume's own index put a Georgius Uroda on folio 6 — folio 6 has now been read end to end, both pages, 1809 and 1810 through to the «Finis», and holds no Uroda at all. The index's *folium* column does not reliably name the leaf. And the doubt carried as «Martini U[roda?]» in the 18 January 1809 entry reads, at five times full resolution, Uzelac — settled, and not a lead. Full record in [notes/karlobag-1807-uroda-gergas.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-1807-uroda-gergas.md).

Ahnentafel 25 is read, 15 September 2026. LIBER DEFUNCTORUM p. 144, image 397: «Udova Maria Antić, supruga pok. Josipa, kći Jurja i Mare Jeličić, R. Katolikinja» — Selce house 29, aged 73, died 6 July 1882, buried the 7th. Widow of ahnentafel 24, daughter of ahnentafel 50, and the age matches a birth of 22 March 1809. The mother is written «Mare» where the tree has ahnentafel 51 as *Margarita* — consistent, not proof, recorded as read. The index does not cover this register, which changes how the rest gets done. A sweep for *Antić* at Selce returns 119 indexed entries and every one is a baptism. The index-with-ARK shortcut that found the Otočac households and the Karlobag baptism in minutes does not work here — this book has to be walked. So the film is mapped instead. 005497948 is 802 images and is not one book: image 200 is a *HISTORIA PAROCHIAE*, the death register runs through the 380s–390s, image 600 is a later Croatian-ruled register, and the waypoint in the URL changes between images. Measured: 384 = 1877, 397 = 1882, 400 = 1883–84 — about 2.7 images to the year. Ahnentafel 24 is read, 15 September 2026. Page 99, image 340, ARK `3:1:3QSQ-G99X-13PY`: «*Letta 850. na 21. 8bre* · house 74 · Preminu Osip Antić težak i Mare kćer · 44 / 6 · *Od kolere*». One entry for two people — Josip and his six-year-old daughter Mare, dead the same day of cholera. The age puts his birth in 1805–06 and the tree says 14 March 1806. His house is 74, not the 29 where his widow died in 1882 — two numbers, thirty-two years and a widowhood apart, recorded as two. The page is a casualty list. *Od kolere* runs down the remarks column; pages 98–99 hold 30 September to 21 October 1850, six deaths on the 10th alone, and seven other Antić deaths in the eleven days before Josip. None is claimed. A false null was nearly written and the page numbers stopped it — 99, [100], 101 across images 340 and 341, so the jump from 21 October 1850 to 21 February 1851 is a real gap in the dying, not a lost leaf. Written up in [notes/selce-cholera-1850.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/selce-cholera-1850.md). The year map was wrong and is fixed. Not 2.7 images to the year: 311 = 1834, 338 = early 1849, 339–340 = Sept–Oct 1850, 341 = 1851–52 — about 1.7 through the 1830s–50s, 2.7 only in the 1870s–80s. Arks in `sources/familysearch/film-005497948-selce-deaths.json`. What is left is row 25: ahnentafel 48 and 49 (1829, ≈ image 302) and 96, 98 and 99 in the 1790s–1800s.

Their son was baptised 1 November 1857, so the marriage falls just before it. Their register names are now known exactly — «Juraj Papa, težak» and «Matia rođena Pilipić» — which is what makes this findable. Not in the index; a film read. The FamilySearch block of 15 September 2026 comes and goes — see row 21. It shut three times and opened three times in forty minutes, and fifteen register entries were read in those windows. This row is not blocked; it is waiting for the next window, with its targets written down. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Tested directly for this row and it is not there. Searches for a Georgius Papa or Papić marriage, for a Matia Pilipić marriage, and for a Papić × Pilipić pair all return nothing. Row 4 remains what it always said it was: a film read, and that is now a measured fact rather than an expectation. Retested against the marriage index that does exist, and it is still not there. Five queries — Juraj Papić × Matia, Georgius Papa × Pilipić, Matia Pilipić × Papa, × Papić, and Papić marriages at Senj and at Selce in the 1850s. Every hit is the 1885 marriage of their SON Josip, where the two of them appear as his parents. The index holds the son's marriage and not theirs. Row 4 is a film read, now tested five ways rather than assumed.

Found 15 September 2026. «Ego Joës Antonius Frankovich Parochus Loci baptizavi Infantem *die 11ma* hujus natam ex *Vincentio, et Oliva Smoyver, vulgò Vukić*, Conjugibus hujus Civitatis, cui imposui est nomen *Oliva Maria*» — Karlobag, page 65, ark `3:1:3QS7-L99C-L998-5`. The month column reads *Ejusdem* under *Martij*, and the next opening begins at 19 March, so 11 March 1781. The tree had only «MAR 1781». It confirms ahnentafel 106 and 107 from the register side — Vicentius Smojver Vukić and Oliva Bacchi were in the tree but had never been put on a page; one line now names the couple and the child together. And it gives a by-name in a parish that had none. «Smoyver, *vulgò* Vukić» — this archive's by-name file was built out of Selce, where the connectors are *aliter · dictus · dicti* and the scribe underlines the pair. Karlobag uses *vulgò* and does not underline. It also explains a name the tree carries without comment: ahnentafel 106 is recorded as «Vicentius Fabiano Smojver Vukić» as though Vukić were part of the surname. It is the household name, and the clerk said so. A film note this corrects. `holdings.psv` gives DGS 5481648 as the 1691–1804 baptisms and [notes/karlobag-films.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-films.md) calls it a 327-image film with «image 220 = Anno Domini 1782». Film 005481648 actually has 596 images and its image 216 is headed «Godine 1845». The 1771–1804 baptisms are under a different waypoint, and the way in was the record index, not the film number. Full record in [notes/karlobag-1781-oliva-smojver.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-1781-oliva-smojver.md).

Bribir 1678 onward — its own parish, filed here as a village of Grižane until the catalogue was read properly. Sv. Juraj kod Senja 1697–1852, not in the parish table at all. The **Karlobag *Stanje duša* beyond the U section. The Grižane *Popis duša* of 1738, a household census a century older than Otočac's. All four are in holdings.psv with their film numbers. The FamilySearch block of 15 September 2026 comes and goes — see row 21. It shut three times and opened three times in forty minutes, and fifteen register entries were read in those windows. This row is not blocked; it is waiting for the next window, with its targets written down. --- 21 September 2026 — two of the four are settled without a film being opened, and the honest headline is that three of them reach nobody. The parish-table half of this row is already done. Bribir and Sv. Juraj (kod Senja) were both added to `parishes.psv` on 14 September, catalogue-verified. What was left was opening the books. Two no longer need opening in the way this row imagined.** * GRIŽANE IS DENSELY INDEXED, and `indexed.psv` was badly wrong about it — it said «one hit; treat as untested, not as coverage». A place-only census returns 200 of 200 in both windows, events 1698 to 1894, and burials as well as baptisms: 104 burials against 96 baptisms in the 1800–1899 window. Two collections, including *Croatia, Primorje-Gorski Kotar, Births, from 1569 to 1901*. The Popis duša of 1738 is still a film read — a household census is not an event index — but everything around it is a search box. * SV. JURAJ KOD SENJA IS A MEASURED ZERO. Probed place-only under all three strings the parish table gives it — *Sveti Juraj*, *Jurjevo*, *Lukovo* — across 1690–1869. Nothing. Filmed 1697–1852 and not indexed, so that one is a sitting or it is nothing. AND THE THING THIS ROW SHOULD SAY OUT LOUD: this archive has no person at Bribir, Grižane or Sv. Juraj. Not one, in any of `bornPlace`, `diedPlace` or `baptisedPlace`. Three of the four books are breadth, not descent — worth opening one day, and not ahead of a book that reaches somebody. The fourth is different. The **Karlobag *Stanje duša* beyond the U section is at a parish where this archive has ancestors on the direct line, and no index substitutes for a household census. That is the one of the four worth a window**, and this row should be read as one target rather than four.

Image 75. Every other *Pagus* in the book was mapped — Ottosaz, Luka, Dubrava, Novoselia, Lieška, Kostelievo Selo, Obilje, Beljue, Šumećica — and this one resisted. Small, but the map is otherwise complete and a hole in it will be noticed. The FamilySearch block of 15 September 2026 comes and goes — see row 21. It shut three times and opened three times in forty minutes, and fifteen register entries were read in those windows. This row is not blocked; it is waiting for the next window, with its targets written down. --- 21 September 2026 — READ, and not identified, which are two different things. The heading resisted because it is not where the other headings are: it is written vertically down the left margin. Read at 4× off the full-resolution scan, under the word *Pagus*: > «Vivõz» — a macron over the *o* — written twice, on two successive openings. A shorter word stands above it, «Sched…» or «Scheiz…», not read out. It is deliberately NOT matched to a modern village. A macron in this hand is an omitted *n* or *m*, so the name may be «Vivonz»; and this book renders a final z as č — «Ottosaz» is Otočac — so the Latin letters are not the Croatian name. A modern identification would be a guess dressed as a reading. The hole is filled to the extent an image can fill it: the letters are on the page instead of the word «unresolved». Naming it wants a gazetteer of the Otočac parish. And this row says «the one village» when there are two. The film map records image 110 as «another village», equally unnamed. The method matters more than the village. The film map told the next reader to «rebuild the rest by opening the thumbnail grid and scraping alt="Image N"». That is obsolete — the whole film enumerates in ONE call, and an image can be read without an ark at all at `<das>/v2/<apid>/dist.jpg`, which is how image 75 was read since no ark for it had ever been recorded. Verified against image 95, whose ark this archive already had. That bears on row 21: finding the image is no longer part of the twelve round trips, and neither is having an ark. [notes/image-75-reads-vivoz-and-the-film-enumerates.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/image-75-reads-vivoz-and-the-film-enumerates.md)

Done 15 September 2026. Three rows were retired, not edited: Ancestry, FamilySearch — Rijeka registers and FamilySearch — Otočac, Stanje duša were *placeholders* — «never», «not yet», «still unread» — and each had been superseded by a real row recording the actual reading. The register was over-reporting what is outstanding: eleven pending, of which three were already done. Seven now, and every one of them genuinely open. The gate that stops it recurring. Every register row now declares a `key` — the body of material it covers — and `tools/searched.py` refuses when a row is *pending* while any other row with the same key carries a real outcome. The key is there because the source *name* cannot do this job: «Senj marriages, 1734–1858» is genuinely unopened while «Senj marriages, 1859–1920» is read end to end, and a gate matching on names would have called the first one stale and sent a sitting to re-read a book that was already finished. A gate that cries wolf is worse than no gate. The same stale claim had spread into prose: `/searched` still named four portals «never once searched» when all four had been opened, and `requests/README.md` and `REPLIES.md` still said the three letters were not yet sent when they went on the 15th. All corrected.

Georgius Uroda and his wife Maria were read out of a register this week and are real. The tree stops at ahnentafel 54 and this archive does not edit the tree — that separation is what the whole project rests on. They are recorded in `sources/found.psv` with ten others, but nothing on the site shows them yet. Whether a found person should appear beside the tree, and how, is a decision about what this archive *is*, not a task.

Answered 15 September 2026, the same day it was sent. Boris Zakošek: «Državni arhiv u Rijeci ne provodi ovako složena rodoslovna istraživanja. Istraživati možete sami u čitaonici Arhiva ili putem internetske stranice familysearch.org» — *the archive does not carry out genealogical research of this complexity; research it yourself in the reading room or on FamilySearch.* It refuses the service without answering about the holdings. The letter asked one question needing no research: whether they hold, or know who holds, the Rijeka baptismal register covering 1 March 1892. That is unanswered, and it matters — the 1892–99 volume is missing from the film and from its Kazalo both, which is why Ljubomir Blažević's baptism has never been placed. The reply points back at FamilySearch, which the letter had already said — quoting the volume list — has been read to the end with a gap exactly there. What is left is the reading room, which is genuinely offered: this is material that can be consulted in person at Rijeka, by David or by a paid local researcher. That is now the only stated route to the 1892 entry, and it is a decision rather than a task.

Sent 15 September to the parish priest. Three deaths after the film stops in 1907 — including Milka Papić, whose 1886 baptismal entry carries a registrar's 1971 annotation citing page 280, line 37 — the January and February 1895 baptisms, and whether a *stanje duša* survives for Senj.

Answered 15 September 2026, the same day, by vlč. Jakov Filipović, župnik. «Matice i arhiva svih župa kojim upravljam je predata u Riječku Nadbiskupiju. Koliko mi je poznato oni su istu predali na skrb Državnoga Arhiva. Budući mi ovdje na župama nemamo dodir sa starijim Maticama, za daljnje informacije kontaktirajte Ured riječke nadbiskupije.» It corrects the assumption this whole line of letters rested on. The reason for writing to parish offices was the Senj archive's instruction plus a belief that *the books the films do not cover are the ones that never left parish custody*. For Krmpote that is simply wrong. The registers went to the Archdiocese, and the priest has no contact with his own older books. Nothing was refused. Juraj Blažević's baptism, his marriage, his death and the Butorac marriage are all still wanted; they are just not in this man's hands, and he said so kindly and at once. The door it opens is row 52.

Sent 13 September, acknowledged and queued on the 14th, nothing since. Gospić holds Otočac and Karlobag — which is to say the Žubrinići and the Urode.

`collections.arolsen-archives.org` refuses this machine. It is the likeliest route to anything on Senj, 1943, and it has never been searched. Ten minutes in an ordinary browser.

Fifteen nodes named with a single capital letter — «M Žubrinić Xubrinich Zubrinic» — were being published as parents of 75 real people and as one another's siblings. A card index typed in as a family. Both the bucket detector and the site's junk filter now refuse them, but all seven archives share one GEDCOM, and nothing here has checked whether the same drawers exist under Blažević, Papić, Prpić or Kalanj. The test is four lines and belongs in the kit rather than in this archive. --- 21 September 2026 — answered, and the answer is two answers pointing opposite ways. THE PROBLEM IS UNIQUE TO ŽUBRINIĆ. A sibling set of three or more in which *every* member is alphabet-shaped and dateless fires once in all 15,643 records: family `@F501632@`, the fifteen, with eighty-two real people published as their children. Every other sibling set holding an alphabet-shaped member has one or two of them among real siblings — Corbett 2 of 11, De Franceschi 1 of 14, Jelcic 1 of 11, Defranceschi 1 of 10, Willemse 1 of 6, Leonesio 2 of 5, D'Arcy 1 of 3. THE TEST IS NOT UNIQUE TO ŽUBRINIĆ, AND THAT IS THE WARNING. This row says «the test is four lines and belongs in the kit». The four-line version has ten real people in its blast radius. The per-node test — one leading capital, a surname, no dates — matches 31 records, and only fifteen are the drawer. The other sixteen are in the sister archives: W Florance Barry · H E Corbett · J C Corbett · J M Willemse · G Norman D'Arcy H W Pretorius · N L S Klaassen · E Riedo Defranceschi · J B D Defranceschi · M K Defranceschi These are people recorded by their initials, which is an ordinary English and Afrikaans convention and not a filing artifact. *H E Corbett* is a man his own family knew by his initials. The remaining six are lone letters with no children at all — stubs that parent nobody and misfile nothing. SO THE RULE IS ABOUT A SET, NOT A NODE. What makes the fifteen a drawer is not that each is a letter; it is that they are entered as one another's siblings and eighty-two people hang off them. A single initial tells you nothing. A family all of whose children are single initials is a filing cabinet. Nothing changed here, deliberately — `buckets.py` and the junk filter already refuse the fifteen, and the other sixteen are outside this archive's surname scope. The measurement and the warning have gone to the session that keeps the kit, which is where this row said the test belongs. [notes/the-alphabet-is-one-drawer-the-test-is-not.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-alphabet-is-one-drawer-the-test-is-not.md)

The index gives the image. FamilySearch's record search returns the image ARK of every indexed entry, so a record opens at the page in one step instead of being hunted down a film — that single fact produced most of this week's findings. A third FamilySearch failure state: a sign-out takes *everything* (503, redirect to the sign-in host) where a rate-limit blip and a session block leave the catalogue answering — written into `tools/familysearch.md`. And `sources/found.psv`, somewhere to put people read out of a register whom the tree does not have; every archive here will hit that wall. Fourth — the register cannot be trusted to retire its own rows. `/searched` here said eleven sources were outstanding when three had been read and closed: each was a «not yet» placeholder left standing because the reading went into a *new* row. Over-reporting what is outstanding is the same lie as under-reporting it, and it costs a sitting that re-opens a finished book. The fix generalises: **every row declares a `key` naming the body of material it covers, and the checker refuses a *pending* row when another row with that key has a real outcome.** The key is the whole trick — matching on the source *name* would flag «Senj marriages, 1734–1858», which is genuinely unopened, against «Senj marriages, 1859–1920», which is read end to end. Every archive that keeps a search register has this bug waiting in it. In `tools/searched.py` here. Fifth — there is a fourth FamilySearch failure state, and it is the worst one. Beside the rate-limit blip, the session block and the sign-out, there is Error 15: a 403 «Access Denied» interstitial on *every* path, laid on the egress address rather than the session. Signing in again does nothing. The diagnostic that tells it apart in one call is to fetch a Deep Zoom `image.xml` for an ark that read successfully before: a blip or a session block leaves the tile edge answering, and Error 15 takes the tiles down too. If the tiles are gone there is no route to a page and the only moves are to wait or to read from a different network. Sixth — a work list cannot see its own omissions, and needs a second list to count against. The kit's `checkworklist.py` gates the rows a list *has*. Nothing gated the rows it was *missing*, and on 15 September this archive had five letters sent and four rows — the missing one being the reply that named three further doors, and therefore the reason three of the other rows existed. Every row present looked fine in isolation. The list was only wrong in what it did not say. `tools/checkcovers.py` here counts the work list against two lists that are already exact: every `pending` row in the search register (which means, in the register's own words, «known, wanted, and not yet got») and every letter in `requests/sent/`. A row declares what it accounts for in a `covers` field, and the build refuses when something outstanding has no row — or when a row points at something that no longer exists, which is how a list goes stale rather than short. It found two real gaps on its first run — Brinje, pending since the 14th with no row, and ahnentafel 11 — and it rejected its own first draft: that draft exempted keys beginning `plan/` as «not really work», and the one row so excused turned out to name two baptisms sitting in a book already open. The exemption was noise-reduction dressed as a rule, and there is none now. Every archive in the estate keeps a work list and a search register, so every one of them can be short in exactly this way, and none of them can currently tell. --- Sixth and seventh, found 20 September 2026, and they are the two that will travel furthest — because they are both about the SAME mistake: trusting the query you sent instead of the answer you got back. SIXTH — A SURNAME IS THE WORST THING TO SEARCH AN INDEX BY, and every archive here searches by surname first. Ahnentafel 52's baptism sat in the FamilySearch index for three sittings under «Piberich» — a form that occurs exactly once in the whole parish, because it is one clerk's reading of one hand rather than a name anybody bore. Six spellings had been tried. No list of guesses ever contains «Piberich». The fix is to stop naming the surname. `q.givenName` + `q.fatherGivenName` + `q.motherGivenName` + a date window + the parish returns the household whatever the clerk did to the family name — the given names are short, common and survive a bad hand, and two parents plus a date is already a near-unique key. It returned ahnentafel 52 and three brothers in one query. Every archive with a parish register and a mangled surname has this waiting in it. SEVENTH — A ZERO IS ONLY EVER «NOT FOUND BY THIS INSTRUMENT», AND A COVERAGE TABLE BUILT FROM ONE INSTRUMENT IS A LIST OF ITS BLIND SPOTS. `sources/familysearch/indexed.psv` recorded seven parishes as NOTHING, each «tested under» two surnames. Re-probed by place alone, naming no surname, three of the seven were wrong — Bribir had 80+ indexed deaths, Selce five baptisms, Ledenice four entries. The Selce row indicted its own method: it recorded a zero «under Antić» and one of the entries it missed is «Margaritam Antich». Run the controls. Karlobag and Senj return 100 of 100 in every window, which is what says the instrument works at all. And read the place string BACK — «Bribir» is two villages 200km apart and the wrong one outranks ours 593 to 7; «Smokvica» returned a village in Slovenia. Neither probe is authoritative alone: the place-only probe returned zero for Krmpote, where the surname probe had found a 1895 baptism the same day. [notes/the-zeros-were-not-all-zeros.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-zeros-were-not-all-zeros.md), [notes/karlobag-the-index-calls-them-piberich.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-the-index-calls-them-piberich.md) --- Eighth, 20 September 2026 — TWO PATHS TO «READ» AND ONLY ONE OF THEM CHECKED. `readings.psv` here had ninety-three rows and eight of them credited nobody: the slug existed in no data file, so `coverage.py` looked it up, got nothing and moved on, and `datecheck.py` did `people.get(slug)`, got `None`, and skipped the row. Nothing crashed and the file looked fuller than it was. All four bad slugs were a nickname the slug generator drops and a hand-written row keeps — `…-MATIJA-pilipic`, `…-VICKO-pilipic`, `…-JOSO-papic`, `…-JURE-antic` — and all four are direct-line ancestors. It went unnoticed because those people were *also* graded read through a citation keyed on the GEDCOM id, so the broken path was masked by the working one. Any archive that joins hand-written evidence to a generated tree has this. The fix is four lines: refuse the build on a row whose key names nobody, and print the nearest real key, because the cause is always a near-miss. In `tools/datecheck.py` here. --- Ninth, 20 September 2026 — CHECK THAT THE SITE'S EVIDENCE IS IN THE FILES THE TOOLS READ, because here it was not. A request to «make sure everything researched is on the site» found the traffic running the other way, and harder. A note written that morning concluded ahnentafel 3's birth date could not be settled because «there is no image yet» — and the page had been read years before and was published on this site, image and all, with the entry transcribed. Her 1907 marriage repeats the date in a second hand. The note had checked the index against the TREE and never against THIS ARCHIVE. Holding `corrections.json` against `readings.psv` found thirteen more, including the marriage of 8 December 1920 — the document the whole archive was built around, with its own page on the site — where ahnentafel 2 and 4 had no readings of any kind. Fourteen rows recovered; `readings.psv` 93 → 111 rows, dates checkable against the tree 19 → 24, all still agreeing. Why it went unseen is the transferable part. `coverage.py` grades a person read from either a reading row or a citation keyed on the GEDCOM id. The corrections carried the citation, so everyone graded correctly while the archive's own record of its work quietly did not hold them. Two paths to a conclusion, and a working one masking a broken one — the same shape as the eight orphaned slugs found the same day, and the third instance in one sitting. `tools/publishedcheck.py`, a report and not a gate: a correction can legitimately have no reading when what it found is a disproof — ahnentafel 20's says «the 1845 baptism is *not* his», and recording that as a reading would assert the opposite. The tool separates disproofs out. [notes/the-evidence-file-under-reported-the-archive.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-evidence-file-under-reported-the-archive.md) And the one thing genuinely missing from the site has been published. `sources/familysearch/indexed.psv` — the table that decides how every register row is planned — was read by no tool and shown on no page. It is now at `/indexed/`, built by `tools/indexcoverage.py`, leading with the two occasions it was wrong. --- Tenth and eleventh, 21 September 2026. TENTH — A FILM ENUMERATES IN ONE CALL, AND AN IMAGE NEEDS NO ARK. Every archive here prices register work as «find the image, then read it», and the finding half is unnecessary. `ark → apid`, `apid/parents → film`, `film/children → every image on the film` as `{apid, name, order}`; then `<das>/v2/<apid>/dist.jpg` is the page itself at native resolution. An image with no ark recorded anywhere is reachable this way — which is how this archive read a page it had never been able to address. Six archives share one set of films; none of them should be scraping a thumbnail grid. See row 21. ELEVENTH — A DETECTOR WRITTEN FROM ONE ARCHIVE'S EXAMPLES WILL FIRE ON ANOTHER'S REAL PEOPLE. Row 15 wanted the «alphabet node» test — one capital, a surname, no dates — promoted to the kit as four lines. Run across all 15,643 records it matches 31, and only 15 are the card drawer it was written for. The other sixteen are in the sister archives and at least ten are people recorded by their initials — *H E Corbett, W Florance Barry, G Norman D'Arcy, N L S Klaassen* — which is an ordinary English and Afrikaans convention. The rule that travels is about a SET, not a node: a sibling set of three or more in which *every* member is alphabet-shaped and dateless. That fires once in the whole estate and would fire zero times in six other archives. A single initial tells you nothing; a family all of whose children are single initials is a filing cabinet. The same shape as `namefold`'s corpus reading prose: a test is only as portable as the examples it was written from.

The styles.css instance cleared — whoever wrote the burger-menu fix has committed or discarded it, and it is gone from the tree. A second instance took its place the same day. `site/package.json` and `site/package-lock.json` now pin the kit at `f55559d` where the last commit here pinned `60aad58`, and `node_modules/` holds the newer one. Nothing in this archive's build script installs anything, so the bump was made by the kit session running `npm install` in this shared tree. It is real work — `f55559d` gives `WorkList.astro` the ring that links each archive's work list to the others — and the kit has already moved one commit past it (`27cfb83`), so the bump is mid-flight rather than finished. Landed by its own session, 15 September 2026 — commit `eb41806` pins the kit at `d1e902e`, past both the version that appeared in this tree and the one after it, and the working tree is clean again. Leaving it alone was the right call: the session that owned the upgrade finished it two commits further on than the state found here. Left uncommitted deliberately at the time. This archive's build passes against the newer kit (check:kit, check:living and check:worklist all green), so nothing is broken; but committing another session's half-landed upgrade would take the decision of *which* kit commit the estate stands on away from the session making it. Every `git add` here stays scoped to `sources/ tools/ site/src/ notes/ requests/`.

Founded on a mistake and kept here rather than deleted. Two sittings went into this book on the assumption it held the family. It does not: Šumećica occupies images 102–106, and every household in them is Dašović, Štabinčić, Orehović, Janežić, Ollar, Kovačić or Orehovački — no Žubrinić, no Domus 2, no Domus 24. The book gave the village map and a clean negative, and it will not give the house. The baptism registers replaced it — that is row 1.

Founded on a mistake. The plan named this «the reachable target» on the strength of a catalogued range of *1815–1896*. The film begins in 1879 and Juraj was born in the 1860s; Krmpote is one film, one item, and there is no other Krmpote book. Ledenice and Novi Vinodolski were checked the same way and hold no candidate. What survives is the letter — row 12.

Founded on a mistake, and it was this archive's own. The morning's plan said *every* entry writes the household into the *Conditio* column. That book does not: its *Conditio* holds one status word — *Confiniarius*, *Decurio* — checked at December 1839 and March 1845, with the place column giving only «Ottoccii». Ahnentafel 10's January 1834 entry is an early-book habit the parish dropped. House numbers are readable for 1846–58 and for the first years of 1834, and not in between.

Error 15, and it comes and goes. On 15 September 2026 every request to `familysearch.org` began returning 403 «Access Denied — Error 15» as a full interstitial. It shut at 04:21, opened at 04:30, shut at 04:45, opened at 04:47, shut at 04:51 and opened again at 05:00. Fifteen register entries were read in those windows, which is the whole of work list row 1. So the lesson is not «wait for it to clear». It is to have the targets — ARKs, dates, parents — written to disk *before* a window opens, so a ten-minute opening is ten minutes of reading rather than ten minutes of searching. That is what `sources/familysearch/otocac-michael-1846-1858.json` is, and it is why row 1 closed on a day when the site was mostly shut. The one-call diagnostic. Fetch a Deep Zoom `image.xml` for an ARK that read successfully before: a rate-limit blip or a session block leaves the tile edge answering, and Error 15 takes the tiles down too. If `image.xml` returns HTML instead of `<?xml`, there is no route to a page — not the index, not the catalogue, not the scans. Signing in again does nothing; the block is on the egress address, not the session. Client IP 168.140.255.212. Reading from David's own browser on his own network would not be behind this address at all. --- A second constraint, and it is the binding one. 15 September 2026: the Browser pane is hidden in the app, and a hidden pane cannot take a screenshot. Every read of a register scan in this archive goes through one — crop, enlarge, look. Without it, three routes were tried and all three failed: * Post the image to a local server. The page cannot reach `127.0.0.1` at all: the in-app browser applies the private-network rule and an `https` page on familysearch.org gets «Failed to fetch», with no CSP report, for both `127.0.0.1` and `localhost`. * Fetch the Deep Zoom tiles from the shell. `image.dzi` answers 403 with a bot interstitial and the tile path answers 401. The tiles want the session, and the session is not something this archive will copy out of a browser. * Carry the image out as base64 through a tool result. This *works at small sizes* — a whole page at contact-sheet scale is about 3 KB and arrived intact. It fails at useful sizes. A 19,891-byte contact sheet came back with exactly the right byte count and the wrong SHA-256: characters substituted, not dropped, silently. Chunking to 76-character lines did not help. So there is a ceiling of a few kilobytes, and a register entry at full resolution is above it. That matters more than it sounds: this archive's own standing rule is that no date, name or house number is recorded from a page-scale view, and it exists because a *19* reads as *1* and a flat-topped *2* reads as *7*. Reading through a channel that can alter bytes without saying so would break that rule in a way nothing downstream could catch. Text is not affected and was used hard. The catalogue is JSON, so row 42's eighty-six film notes came out whole and were verified by SHA-256 against a second fetch before being written down. That is the shape of the workaround: anything that can be checksummed can come out; a photograph cannot. Showing the Browser pane again puts rows 4, 6, 7, 24–35 and 54 back within reach. --- 20 September 2026 — and the three states can be told apart in one call. The notes above describe each failure separately and each test costs a page load. There is a cheaper one: fetch a catalogue item and read the status code, nothing else. | | | |---|---| | 200 | signed in and working — the body is JSON with the film notes | | 401 | signed out. Stop. The account holder must sign in. Waiting does not fix an expired session, and credentials are never this archive's to type | | 403 | blocked — the body carries *«Access Denied … Error 15»*. This one comes and goes: on 15 September it shut and opened six times in forty minutes and fifteen register entries were read in the open windows | The status code alone separates them, and the distinction is the whole decision — a 403 means wait a few minutes and try again, a 401 means stop and ask. Telling them apart by loading pages costs a minute each time. One refinement to what was written on the 15th. That sign-out gave 503 on the personas endpoint; this one gave 401 on the catalogue endpoint. So the *code varies by endpoint* and is not by itself the signature. What is invariant is that a signed-out session fails on the catalogue too — and the catalogue is the cheapest thing to ask. Tested today, and the answer was 401. FamilySearch has been signed out since at least 17 September. It blocks work list rows 4, 6, 7, 24, 25, 26, 27, 29, 31, 32, 33, 35, 50, 51 and 54 — every register row this archive owns. Everything that does not need it carried on: rows 46 and 57 both moved today. --- 21 September 2026 — two thirds of the cost this row measures was finding the page, and that part is gone. This row prices a full-resolution read at «about twelve round trips per page», and treats an ARK as the thing you must have before you can start. Both were true and neither is any more. A FILM ENUMERATES IN ONE CALL. Fetch any known image's ark with `Accept: application/json` and take its `apid`; `GET <das>/v2/<apid>/parents` gives the film's parentApid; `GET <das>/v2/<parentApid>/children` returns every image on the film as `{apid, name, order}` — 384 of them for the Otočac film, in a single request. `das` is `https://sg30p0.familysearch.org/service/records/storage/dascloud/das`. AND AN IMAGE READS WITHOUT AN ARK AT ALL, at `<das>/v2/<apid>/dist.jpg` — a plain JPEG at native resolution, 5793 × 4043 for the Otočac book. That is how image 75 was read, and no ark for image 75 had ever been recorded. The old route could not have started. Verified before it was relied on: image 95's apid from `/children` matches the ark this archive already held for image 95, so the ordering is checked and not assumed. What this leaves. The remaining cost is magnification, not navigation: the browser pane cannot crop, so a region is cropped and rotated onto a canvas in the page and screenshotted — which is how a *Pagus* heading written vertically down a margin was read at 4×. Three or four round trips, not twelve, and the film map's old instruction to «scrape the thumbnail grid» is obsolete. [notes/image-75-reads-vivoz-and-the-film-enumerates.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/image-75-reads-vivoz-and-the-film-enumerates.md)

Done, 15 September 2026. `found.psv` had been recorded since September and rendered nowhere. It is now at /found/, in the nav under Evidence, with the direct-line ancestors separated from the collaterals and every citation printed. And publishing it shortened it. Writing the page meant re-testing the claim it rests on — *«not in the family tree»* — which is row 47, and four rows came out: Thomas Žubrinić, Luce Žubrinić, Helena Orešković and Ivan Defrančeski, who is Hedviga's husband. Twenty-two became eighteen, which is the number this row was written about. The page says what the list is for and what it has been wrong about, because a reader who does not know that «not in the tree» once meant «not among the published people» cannot judge the rest of it.

Corrected 15 September 2026 before it was begun: this is a walk, not a lookup. The row was written asserting the Karlobag deaths could be reached through the FamilySearch index. They cannot — row 43 tested four surnames and every Karlobag hit is a baptism; the 1717–1861 range that suggested otherwise is baptisms throughout. The highest-yield target in the archive, and it is already mapped. `targets.json` says the Karlobag death books reach ten direct ancestors — ahnentafel 52, 55, 104, 105, 107, 108, 110, 111, 212 and 220 — more than any other book here. Four of them have exact death dates inside the 1774–1804 volume: ahn 212 Josiph Smojver Vukić 16 May 1776, ahn 104 Jacob Pilipić 10 April 1778, ahn 220 Nicolaus Gerkacs 30 October 1783, ahn 105 Catherina Luckinić 29 May 1790. Ahn 108's burial of 10 September 1791 is already read out of this same book. And it needs no walking. `holdings.psv` records the register at images 356–409 and its KAZALO A–Z at 410–434, with the letter positions measured — P.R at 428, R.S 429, S.T 430, T.U 431, U.V 432, V.X 433. Look up *Pilipić*, *Smojver*, *Gerkacs* and *Luckinić* in the index, take the folium, convert with the known folio map (left pages EVEN, folio 62 = image 388), and read four ancestors in an afternoon. What the index page actually looks like, now it has been opened. Image 428 is P on the left page and R on the right — the big letter is written at the top centre of each *page*, three column-groups to a page, each *Cognomen · Nomen · dies · folium*. The entries are ordered by folio, not alphabetically within the letter: folios 104, 104, 105, 105, 105, 106, 106, reading Perpich · Perpich · Perpich · Perpich · Potocgniak · Pekkas · Potocgniak. So a surname is found by scanning the letter's pages, not by jumping to it. Folio converts to image as image = 388 + (folio − 62) ÷ 2, left pages even. A find already, from the one index page opened: *«Pekkas Magdalena, 5 Augusti, folio 106»*. Pekass is three of the direct ancestors with no document — ahn 213, 426 and 427 — and this is the first time the surname has appeared in a register here at all. --- Begun and half done, 15 September 2026. The KAZALO works. Ahnentafel 104 is read — folio 17, image 364, ARK `3:1:3QS7-L99C-GSBG`: > «10: Aprilis · D. Jacobus Pilepich Capitaneus portûs, munitus omnibus Sacramentis per M.R. D. Andream Pintor Capellanum Loci, spiritum Deo reddidit, … in propria Crypta in Parochiali Ecclesia existente, per me Joannem … tumulatus est» Three marks of standing in one entry, and this archive had none of them. *Capitaneus portûs* — captain of the port, a crown office in a Military Frontier harbour. The *D.* is *Dominus*, withheld from the *Decurio* and the *infans* on the same opening. And he is not in the parish cemetery: he is in his own crypt inside the church. No age is given. [The note](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-1778-jacobus-pilepich.md). The route, for the three still to do. Image 427, right page, is the letter P — three dense columns, ordered by folio, not alphabetically. «*Pillepich Jacobus · 10 Aprilis · folio 17*» sits in the third. Then folio → image, measured directly: image 364 = folios 16|17, image 365 = 18|19, so `image = 356 + folio ÷ 2`. That disagrees with `holdings.psv` by one — it puts folio 62 at 387 where the file says 388. Read the folio number printed at the top of the page, every time. Ahnentafel 212 is read too — folio 11, image 361, ARK `3:1:3QS7-L99C-G3WG`: «*16 ejusdem [Maij 1776] · Josephus Smojver dictus Vukcich, annorum 45, munitus omnibus [Sacramentis] … in Domino obijt, atque die 17 … in Coemeterio Parochiali tumulatus*». The tree's date to the day, and *annorum 45* agrees with its 1731. He is not under S. He is under V — «*Vukich — vide Smojver — 16: Maij — folio 11*». This KAZALO indexes the household by-name as a headword and cross-references it to the surname, and does the same on the same page for «*Verbanovich … vide Urbanovich*». A surname missing from its own letter is not absent from the book. For a family already known here as «*Smoyver vulgò Vukić*» (1781), that is the difference between an ancestor and a null — and the 1776 entry uses a different connector, *dictus*. Ahnentafel 220 is read — folio 39, the right-hand page of image 376, ARK `3:1:3QS7-L99C-GSBH`: «*30 8bris · Nicolaus Gergacs annorum 66, munitus omnibus sacramentis obiit … in domo filij sui Antonij, et die 31: ejusdem … sepultus est in Cœmeterio … Laterali ex parte dextra Ecclesiæ Parochialis hujus*». The tree's date to the day, *annorum 66* agreeing with its 1717 — and he died in his son Antonius's house, a Gergacs this archive has never published. Two things the sitting taught. * The KAZALO's date can be the burial. It says «*3i 8bris*»; the register says died the 30th, buried «*die 31: ejusdem*». That is exactly what Find a Grave does — a cemetery records the interment, a parish the death — and a parish index does it too. Use the index date to *find* an entry, never to record one. * The folio arithmetic drifts; the index does not. Interpolating from folio 17 put folio 39 on image 375, one leaf early, on a page that ends at 18 June 1783. The printed number settled it in one crop. Interpolate, then read the number at the top of the page. Letter positions measured: G 418 right (419 left is G and empty), P 427–428, R 428, S 429, T 430, V 432. Ahnentafel 105 is read, and she closes the row — folio 63, image 388, ARK `3:1:3QS7-L99C-G3QY`: «*29 ejusdem [Maij 1790] · Catharina annorum 70, vidua defti Jacobi [P]ilepich, obiit in Domino munita omnibus sacramentis, atque die 30 sepulta est*». The tree's date to the day — and what identifies her is her husband's name, not her own: she is the widow of ahnentafel 104, the port captain, dead twelve years before her. The age disagrees: 70 in May 1790 puts her birth about 1720 where the tree says 30 March 1727. Recorded, not resolved. She is not in the KAZALO's L, and the whole of L was read to prove it — folios 1–73 and 76–102, group 3 blank, image 423's L page empty. Five Luckinić entries, none Catharina. A widow is indexed under her husband's surname, which is the third time today this archive has met that rule: «*Vukich vide Smojver*», the Senj stones carving a married woman under her husband's name, and now the KAZALO. Two more by-name headwords fell out of the search — «*Ljubicich dictus Luxich ut ibi*» and «*Locatelli, vulgò Zessich*». Year map for this volume, measured: folio 17 = 1778 · 39 = 1783 · 63 = 1790 · 74 = 1793 — about 3.5 folios to the year. From image 376 to 394 it is exactly two folios per image; between 364 and 376 it is not. All four ancestors in this book are read. The next one is row 24, the *Umrli 1804–1861* on the same film, whose own alphabetical index a previous sitting already found at images ~168–192. And the index has put Pekass on the table. Five entries — *Pekas Anton* folio 1, *Blasius* folio 2, *Maria* folio 5, *Joanna* folio 10, *Pekkas Magdalena* folio 106. Pekass is ahnentafel 213, 426 and 427, three direct ancestors with no document at all, and the surname had never appeared in a Karlobag register here. These are index lines, not readings — no page has been opened.

The other half of row 23's ten. Ahnentafel 52 (Josephus Pilipić, d. 26 June 1828), 55 (Maria Gerkacs, d. 6 August 1834 — the entry already quoted here as «uxor piscatoris Georgii Uroda», but never read out in full), and 111 (Lucia Tomljanović, d. 19 September 1809) all fall in the 1804–1861 volume, which has its own index on the same film. Do it after row 23, with the index technique already proved. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Karlobag deaths are not indexed. The baptisms are — two collections, 1733–1803 and 1805–1858 — and the deaths are not, under any collection. Ahnentafel 52, 55 and 111 are a film read, exactly as this row planned, and the volume's own KAZALO is still the instrument. A search for Lucia Tomljanović at Karlobag returns Smiljan and Krivi Put baptisms and nothing from the death book. And the coverage table was then crossed against the people who still need a document — with a flat answer. Not one of the 24 direct ancestors still without a document is reachable by the baptism index. Every one is either in a parish with no baptism index at all (Selce, Krmpote, Smokvica, Montani Stanište), or born before the index's coverage begins (Vitus Sekula, 1787, against an Otočac index that starts at 1839), or has no year and no parish — which is seventeen of the twenty-four, and is row 34's population. That is why this archive stalled on exactly these people, and it is worth stating plainly: the index was decisive for row 54's Karlobag walls because Karlobag's is the only index that reaches back to 1733. Nowhere else does it reach the eighteenth century at all. One live thread came out of the crossing. Ahnentafel 210 Vicentius Luckinić and 211 Maria had no place, no date and no document. Their daughter is the anchor: ahnentafel 105, Catta Luckinić, was born 30 March 1727 at Karlobag to Vicentius and Maria, which places the household out of the tree alone. The index adds two children — Mikola, 9 December 1719 and Mare, 13 April 1725, both entered as *«Vicka Luckinicha»* — eight years and two years before Catta, same rare surname, same parish, same father's given name. They are in `found.psv` as candidate siblings and not as proof. Both entries name no mother, so nothing in them distinguishes one Vicko from another. And Catta's own 1727 baptism is not in the index: the entire Luckinić holding at Karlobag is seven entries, which is the best single measure of how partial this indexing is. --- 20 September 2026 — and the last Karlobag households came out, with one that looked like a contradiction and was not. The parents-query on ahnentafel 54 × 55 returned four children across three years: | | | | |---|---|---| | 15 February 1810 | Margaritha Maria Uroda | «JOANNE Uroda × Maria» | | 23 April 1810 | Mathiam Georgiam Uroda | «GEORGIO Uroda × Maria» | | 25 August 1811 | Joannem Bartholomæum | «JOANNE Uroda × Maria» | | 24 August 1812 | Dominicum Bartholomæum | «GEORGIO Uroda × Maria» | 23 April 1810 is ahnentafel 27, so the February child cannot be her mother's — nine weeks is not a pregnancy — and for an hour this looked like the index contradicting the tree about ahnentafel 54's own name. It is two brothers, and this archive already held the fact that separates them. Ahnentafel 54 was baptised *«Joannis Georgij»* — two given names — and is named alive in 1834 as *«piscatoris GEORGII Uroda»*: he went by Georgius, and married Maria Gerkacs in 1807. His brother, baptised Joannis in 1784, is already in `found.psv` with the note *«married Maria filia Nicolai Barletich 10 Feb 1806»*. Two Uroda brothers, married thirteen months apart, both to women called Maria, both fathering at Karlobag in 1810 — and the only thing telling them apart in the index is which of his two given names the priest used. Dominicus Bartholomæus 1812 is ahnentafel 27's brother; Margaritha 1810 and Joannes 1811 are her cousins. All three in `found.psv`, each saying which household and why. And the by-name reached a third generation. Ahnentafel 212's death entry reads *«Josephus Smojver dictus VUKCICH»*, so the index's *«Josepho Vukich × Antonia»* is him: Heva, 10 April 1766 and Joanna, 16 May 1770, two more candidate daughters of ahnentafel 212 × 213. Ahnentafel 212, 106 and their children are all filed under Vukić as often as under Smojver — and none of them would have been found by querying the surname alone. --- 20 September 2026 — the Otočac figure in this row is wrong, and the correction widens the window by nine years. This row rules ahnentafel 84 out as «Vitus Sekula, 1787, against an Otočac index that starts at 1839». Measured by place-only census rather than by surname, the Otočac index begins in 1830: | births | indexed entries | |---|---| | 1700–1799 | 0 | | 1800–1829 | 1 (a stray at Dabar, 1816) | | 1830–1834 | 83 | | 1835–1837 | 200 | The verdict on ahnentafel 84 stands — 1787 is nowhere near 1830, and there is nothing at Otočac before 1800 under any instrument. What changes is the boundary, and `indexed.psv` now carries the measurement instead of the estimate. And the nine years reached one person. Ahnentafel 21, Maria Catharina Orešković, born 8 September 1835 at Otočac — inside the new window, and she had no reading of her own although both her parents had been credited from her baptism since it was read. Fixed, and the check that found it is written up in [notes/eight-readings-that-named-nobody.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/eight-readings-that-named-nobody.md).

Row 3 read ahnentafel 25 (1882) and ahnentafel 24 (1850, of cholera). Ahnentafel 48 and 49 are now read too, both on image 304, ARK `3:1:3QS7-L99X-1QQL`: > «*Die 13ᵃ Martii 1829* · house 51 · *Joannes Antich Papa. obiit Annorum 64*» > > «*Die 15ᵉ Julii 1829* · house 56 · *Margarita uxor Joannis Antich <u>Natanchii</u> obiit Annorum 52*» Both dates are the tree's to the day and both ages fit its birth years, 1765 and 1777. The houses differ and so do the by-names — his *Papa*, hers *Natanchii* and underlined. One household with two names, a widow moved to a child's house, or two different couples: [the note](https://github.com/daviddef/TheBlazevic/blob/main/notes/selce-antic-1829-papa.md) takes none of the three. The marriage, or the Selce *Status Animarum*, would settle it. And «Papa» is question 5's word — here fifty-six years before the Senj «*Juraj Papa*» and behind a different surname. Not a link between the families: **evidence that *Papa* was a by-name in circulation, and therefore weak evidence of descent wherever it stands alone. Three direct ancestors are left in this book: ahn 96 (Joannis Antić, 27 Feb 1793), 98 (Pauli Lončarić, 29 Sep 1811) and 99 (Anna Papić, 24 Aug 1807). `targets.json` scores the book 19 reachable of 19 — the only one in the archive where every reachable person is a direct ancestor or their sibling. The year map is measured now, and it runs back as well as forward: 302 = 1828, 304 = 1829, 311 = 1834, 338 = early 1849, 339–340 = Sept–Oct 1850, 341 = 1851–52, 384 = 1877, 397 = 1882 — about 1.7 images to the year here. 1811 lands near image 273, 1807 near 266, 1793 near 242, all far outside the measured range, and the film changes books more than once. And read the printed page number before writing any negative — it is in the top outer corner of the right-hand leaf. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed**, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Selce is not indexed at all — nothing, under Antić, Papić, Jelličić or Lončarić, for any event. The three remaining ancestors are a film read and there is no shortcut, which is worth knowing before a sitting rather than during one. Ahnentafel 96 (d. 27 Feb 1793), 98 (d. 29 Sep 1811) and 99 (d. 24 Aug 1807) all carry exact dates, so the film can be opened at the year and walked. --- 20 September 2026 — Selce is not unindexed, and this row was planned against a table that said it was. `indexed.psv` recorded NOTHING for Selce, tested «under Antić, Papić, Jelličić and Lončarić». A probe that names no surname at all — place plus a date window, reading the returned place string back — finds five baptisms, 1803–1844. One of them is an ANTIĆ, which is the surname the zero was recorded under: «Margaritam Antich, 7 November 1844, father Josephi Antich, mother Margaritæ», ark `1:1:QKMJ-MSB9`. She is not ahnentafel 12's sister, and it is worth saying why rather than leaving it to look like one. Ahnentafel 24 IS a Josephus Antić of Selce — but his wife is Maria Jelličić, not Margarita, and he died in 1850 with her outliving him to 1882, so there was no second marriage. A different household. She is in `found.psv` to be ruled out, because the next reader will meet her first. What this row should take from it: five entries is not a register, and ahnentafel 48 and 49 were read off the film for good reason. But the parish is not a blank in the search box, and the three ancestors left in the 1790s and 1800s are worth one place-only probe each before the next sitting is booked. The place string is bare «Selce, Croatia» with no county — which Selce these five are in is not yet confirmed. [notes/the-zeros-were-not-all-zeros.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-zeros-were-not-all-zeros.md)

This row was written wrong on 15 September and is corrected the same day. It claimed «seven direct ancestors still unread», taken from the `ahn` list in `targets.json` — but that column names the ancestors a book could serve, not the ones still unread. `coverage.json` is the column that answers the question, and it says seven of the eight are already read: ahnentafel 13, 26, 27, 52, 54, 55 and 106 all have documents. Only ahnentafel 105 is left — *Catherina «Catta» Luckinić*, born 30 March 1727, which falls in the 1691–1770 volume. And it is a walk, not a lookup. The index was searched under *Luckinić · Lukinić · Luchinich* and returns one Karlobag hit across 900 results — a *Mariu Luketich* of 1756, not her. The 1691–1770 book is not meaningfully indexed for this surname. A lesson for every other reading row on this list: check `coverage.json` before planning a sitting, not `targets.json`. The first says who lacks a document; the second says which book would reach them if they did. --- 20 September 2026 — this row was wrong in both halves, and the correction is a method. It has stood since the 15th saying ahnentafel 105 is the last unread Karlobag ancestor and that reaching her «is a walk, not a lookup», the 1691–1770 book being «not meaningfully indexed for this surname». She was not unread. `catherina-catta-luckinic` has held a direct reading — her own death of 29 May 1790, «Catharina annorum 70, vidua defti Jacobi [P]ilepich» — and coverage has graded her `read` accordingly. The row was reporting a gap that had already closed. And the book is indexed for this family after all. Ahnentafel 52's baptism, 11 March 1768, is in the record index — under «Josephum PIBERICH», father «Jacobi Piberich», mother «Catharina», ark `1:1:QKMX-YX3G`. «Piberich» returns exactly one record in the whole parish, so it is not a surname but one clerk's reading of one hand. The same couple appear three more times, as Pillopich (Simon, 21 Oct 1758), Pillepich (Franciscus, 28 Jul 1762) and Pillepich again (Simolicianus, 1 Mar 1765) — all three now in `found.psv` as brothers of ahnentafel 52. Pillepich is this family's own attested spelling, from ahnentafel 104's death entry that this archive read itself: «Pillepich Jacobus *Capitaneus portûs*». What found them was dropping the surname. Six spellings were tried across three sittings and none was «Piberich», because no list of guesses ever would be. `q.givenName` + `q.fatherGivenName` + `q.motherGivenName` + a date window + the parish returns the household whatever the clerk did to the family name. [notes/karlobag-the-index-calls-them-piberich.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-the-index-calls-them-piberich.md) Filed `INDEX ONLY` — no image opened, and the index carries no surname for the mother. Catta's Luckinić still comes from the tree. --- 21 September 2026 — «a walk, not a lookup» is now a lookup with three steps, and the arithmetic that was missing is written down. This row reaches ahnentafel 105 through the FamilySearch *record* index and correctly finds nothing: seven Luckinić entries in the whole parish, none hers. The book has an index of its own, and nothing had connected row 54's waypoint numbering to the film. DGS 005481648 has 596 images. WAYPOINT IMAGE N = FILM IMAGE N + 269 Verified on two independent ARKs before use: waypoint 250 = film 519, waypoint 310 = film 579. So row 54's «image 310 is inside the Kazalo» is film 579, and the Kazalo runs about film 569–596. What the Kazalo is, read at 4×: organised by initial letter then chronologically, five columns to an opening, each entry surname · given name · day and month · FOLIO, and no year at all. Film 579 is the I/J section; film 581 carries L, at folios 109–113. So the remaining work is bounded: scan the L column for the block whose folios fall in 1727 — interpolating from 109–113 — take the folio off her line, open that folio. What stopped this sitting was the length of that scan, not reachability. Every page above was reached with no ARK at all, by film enumeration and `<das>/v2/<apid>/dist.jpg` — film 581 has no ark recorded anywhere in this archive. [notes/the-karlobag-kazalo-is-reachable-and-here-is-the-arithmetic.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-karlobag-kazalo-is-reachable-and-here-is-the-arithmetic.md)

Corrected 15 September 2026, the same way row 26 was. This row was written citing ahnentafel 10 and 21 from the `ahn` column of `targets.json` — which names who a book *reaches*, not who is *unread*. `coverage.json` says both are already read, as are ahn 11 and 20. So the value here is breadth, not the direct line: `targets.json` scores the book 135 people, 91 reachable, second only to Novi Vinodolski. And it pays twice, because this is the book whose *Conditio* column produced the seven Šumećica and Prozor households of row 1 — every entry read gives a person *and* a house number. It is index-led, which is what makes it cheap: Otočac is indexed for baptisms (165 in the sample), and that is how row 1 was done. Do it together with row 51, whose two targets — Lucca 1853 and Marija 1858 — sit inside this same book. The direct-line work at Otočac is in row 28, not here. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Otočac baptisms are indexed, and by TWO collections rather than one. *Croatia, Church Books* carries 1839–1846; Austria, Küstenland, Births, from 1536 to 1919 carries 1848–1858 — and a searcher who queries only Church Books, as this archive has been doing, misses the whole second half of the book. That is most of this row's 91 reachable people, and they are a search box rather than a walk. --- 20 September 2026 — the households, searched by parents. Ahnentafel 42 × 43, Marcus Orešković and Lucia Sekula, are indexed under both surnames — and the tree already knew, carrying two of their children as *«Orešković Oreskovic Sekula»*. A searcher who queries one surname finds half the children. Two are new: Michael, 26 September 1841 and Paulus, 29 June 1846. Paulus is a twin. *Petrus*, whom the tree has, was baptised the same day to the same parents, on a different ark. Row 40's duplicate detector has a *«same day, same parents»* tier it treats as near-conclusive for one person entered twice — here it is two boys, and what separates them is two given names on two arks. And the sexes are swapped. The tree has *Francisca* in 1840 and *Franciscus* in 1844; the index has Franciscus 1840 and Francisca 1844 — same years, same parents, opposite sexes. One is wrong and this archive does not know which. Ahnentafel 26 × 27, Vicentius Pilipić and Matia Uroda — two more the tree lacks: Maria Catharina, 25 January 1834, four years before ahnentafel 13, and Georgius, 18 November 1841. And one household the index cannot reach. Ahnentafel 40 × 41, Toma Žubrinić and Marija Findrić: their child was born in 1811 and the Otočac index begins at 1834. The query returns forty other Žubrinić households and not theirs. That is coverage, not spelling, and no surname variation fixes it. --- 20 September 2026 — the Otočac index has a hard ceiling, and it is this row's own book. A place-only census, naming no surname: | births | entries | | births | entries | |---|---|---|---|---| | 1700–1799 | 0 | | 1855–1859 | 200 | | 1800–1829 | 1 | | 1860–1864 | 0 | | 1830–1834 | 83 | | 1865–1874 | 0 | | 1835–1837 | 200 | | 1875–1879 | 0 | The index runs 1830–1858 and stops — which is *this row's book*, «Otočac baptisms 1834–1858», and nothing after it. So the index and the film cover the same ground, and this row's «index-led» plan is sound for everything inside it and useless for everything outside. It also begins four years EARLIER than the film, at 1830 rather than 1834. Four years of baptisms are searchable that nobody can open on the roll this archive holds. And it is what puts ahnentafel 5 out of reach — Tereza Žubrinić, born 29 November 1867 at Otočac, Hedviga's paternal grandmother, with only a gravestone. Outside the index and outside the film, and so with the post-1894 Senj cases: the parish office or nothing.

The best direct-line target at Otočac, now that row 27 is corrected. Two ancestors are unread and both have exact death dates inside the filmed 1834–1858 book: * Ahnentafel 40, Toma Žubrinić, died 4 January 1838 — and the tree places him at *Šumećica 2*. * Ahnentafel 41, Marija Findrić, died 29 July 1849. Ahn 40 is the one that matters beyond himself. Šumećica N° 2 is the house where row 1 found Nicolaus Žubrinić baptising children in 1856 and 1858. A death entry naming that house would tie the direct line to a house number that is currently only inferred — which is the whole of question 4. It is a walk, not a lookup. Otočac is indexed for baptisms only — 165 in the sample and not one death — so the index cannot be used here (row 43). The deaths sit on DGS 5482657, the same film as the 1846–58 baptisms already read for row 1, so the film is known ground even though the section is not. Unblocked and opened, 15 September 2026 — after a sitting lost to the film number. `…/search/film/005482657?i=N` reports 483 images; the Otočac book is 218, and past that the traversal crosses into Pakrac, which is why four sampled images were all baptisms and a first conclusion said the deaths were not filmed. They are. Waypoint `9RK1-Y4C:391644801,391674001`; image 100 is 1835, 165 is 1849, 200 is 1855. ARKs in `sources/familysearch/otocac-deaths-1834-1858.json`. It is the right instrument. The far-right column is the running entry number — it restarts at 1 under each ruled year heading — and the entry text carries the village and house — «*Confiniarii e Prozor N° 40*». Exactly what question 4 needs. Ahnentafel 41 is read too — image 167, page 153, entry 53. «*Leta 1849 dne 29 Srpnja preminula je · Maria Žubrinić, Udovica · 69 · Slaboća starosti po dokazu Lečnika · u župničkoj grobnici*». ARK `3:1:3QS7-899C-PVHR`. The date to the day, the age against the tree's 1780, and the widowhood all agree — Toma died eleven years earlier. But no house and no husband are named, so her entry is not anchored to Šumećica N° 2 the way his is, and this row does not pretend it is. And a correction, recorded because it is the useful part. An earlier pass today read page 151 as ending *28 Srpnja* and page 152 as opening *9 Kolovoza*, and concluded no death was recorded on 29 July 1849 at all. Every month in that sentence was wrong: 151 runs **4 *Svibnja* (May) to 8 *Lipnja* (June), 152 runs 9 *Lipnja* to 14 *Srpnja* (July)**, and the «*Kate, žena Mate Žubrinića*» entry is 7 May. July had not been reached. The entry numbers did run continuously — which is what made the wrong conclusion feel safe. A continuous sequence proves no page is missing; it proves nothing about which months are on it. Written up in [notes/otocac-marija-findric-1849.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/otocac-marija-findric-1849.md). Ahnentafel 40 is read — image 117, page 45, entry 3 of 1838. «*1838. Januarii die 4ta · Thomas Xubrinich, Inq: ~~Ottoèsanq~~ Centu-rio Confiniarius, ex Pago Shumcchicza N° 2 · An. 70 · Senectutis Debilitas*». ARK `3:1:3QS7-L99C-PVR4`. The tree's date, the tree's house and the tree's birth year are all corroborated at once, for a man who had no media at all. He does not move into house 2 from anywhere — he dies in it, eighteen years before his grandson's children are baptised out of it, and the «he moved / it is another Nicolaus» pair is no longer level. Written up in [notes/otocac-toma-zubrinic-1838.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/otocac-toma-zubrinic-1838.md). And the book turns out to be worth more than two ancestors. The Otočac *baptisms* stop writing the household into *Conditio* from about 1836 to 1845; the deaths write the village and house straight through those years. This book is the house-number source for the window the baptisms cannot reach.

94 people, 73 reachable, carrying ahnentafel 3 and 6 — the closest generations to Seka herself that are still unread. The Senj marriage books have been verified end to end, so the parish is known ground; the baptisms have only been sampled. Note the hard edge: Senj births stop at 1894, which is what makes questions 1e and 5b unreachable (row 33 of the register's «unreachable» list) and why the Senj parish letter was written. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Senj baptisms are indexed by THREE collections over one parish — *Croatia, Lika-Senj, Births* (1692–1861), *Croatia, Church Books* (1735–1857) and *Hungary, Croatia-Slavonia, Births* (1807–1851). The span reaches back to 1692, which is forty years earlier than this row's book is filmed from. Ahnentafel 3 and 6 are lookups. The marriages are not indexed, which is row 4. And the record that disproved the first coverage table is this family's own. *Hungary, Croatia-Slavonia, Marriages, from 1678 to 1927* holds the Senj marriage of 4 November 1907 — Zvonimir Vukelić × Milka Papić, question 1c — with both sets of parents, `1:1:X9D5-DTRW`. His baptism is indexed too, `1:1:X9DP-3Y7X`, «baptised 25 February 1883 at Senj, Petar Vukelić × Lojza Pulanić», confirming to the day what this archive read off the film and confirming that Lojza and Vjekoslava are one woman. And it disagrees with this archive about her surname. Corrections records that at 5,812 pixels the 1907 row reads *«† Vjekoslava Putnić»*, with the surname left unsettled. The index says PULANIĆ, twice, in two different records. Those are not variant spellings — *tn* against *lan*. It is not settled in the index's favour, and today is the reason why. The same indexing names the wrong mother for ahnentafel 106 — *Marie* where the register plainly reads *Antonie*. An index transcription is not evidence against an image. What the index adds is that it is two transcriptions from two different entries, which makes the question live again without answering it. [notes/senj-1907-vukelic-index.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/senj-1907-vukelic-index.md) --- Searched by PARENTS rather than by name, which is the query that finds a household — and it returns five children of ahnentafel 6 and 7. | Child | Index | Tree | | |---|---|---|---| | Marija | born 11 Aug 1885 | 11 AUG 1885 | agrees to the day | | Milka Lucia | born 9 NOV 1886, bapt. 17 Nov | 9 OCT 1886 | DISAGREES — the month | | Anica | bapt. 3 Jul 1888 | 25 MAY 1888 | no birth day in the index | | Josip | bapt. 1890 | 18 MAR 1890 | no day | | Božica Tonka | born 9 Dec 1891 | 9 DEC 1891 | agrees to the day | Milka Lucia Papić is ahnentafel 3 — Hedviga's mother. The day of the month matches and the month is one step out, which is the shape of a transcription slip rather than of two children. Two things sharpen it: the index agrees with the tree *to the day* on two of the other four, so it is not careless about this household; and a baptism eight days after a birth is ordinary here, where thirty-nine days would not be. It is not settled from the index, for the reason this whole day keeps producing: the same indexing names the wrong mother for ahnentafel 106, and gives Zvonimir Vukelić's mother a surname the archive read differently off the image. Her own row is deliberately NOT in `readings.psv` — the four uncontested children are, marked index-only; writing hers in would assert the index over an image that has not been opened. And it is reachable. Senj births are filmed 1878–1894, so 1886 is inside the film. One page settles the archive's second-most-important birth date either way. [notes/senj-1886-milka-baptism-index.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/senj-1886-milka-baptism-index.md) Question 5b is now closed from both directions. The index holds five of this couple's children and stops at 1891 — narrowed three ways, the 1895 sisters are not there. Senj births are filmed only to 1894. Franjka and Šulka are in neither the book nor the index, which is what row 29 always said about 1894 and is now tested rather than inferred. Tested the same hour against the archive's own images, and the test went against the hypothesis. The right test does not use the index at all: `sources/readings.psv` records dates read off the pages themselves, and if the tree systematically lost Roman-numeral strokes, the image reads would show it too. Seventeen dates could be held against the tree. All seventeen agree exactly — baptisms and deaths, 1754 to 1886, across Karlobag, Otočac, Selce, Senj and Krmpote. Not one month disagreement, not one day disagreement. And the one that looked like a third case was this archive's own error. The audit threw up `ivan-blazevic` — tree 25 July 1882, citation 26 Aug 1882 — the pattern exactly, and against an image rather than an index. It was the citation that was wrong. [notes/reading-2026-09-09.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/reading-2026-09-09.md) records the entry as *«1882, 25 / 26 srpnja»*, and srpanj is JULY. Somebody rendered the Croatian month as August when writing the row. The tree was right and this archive's data file was not. Corrected. So the weight has moved the other way. The tree has a perfect record against every image read here, and both month conflicts are tree against INDEX — the party already caught twice today naming the wrong person. This note leaned towards the index this morning on a thirty-nine-day gap between birth and baptism. That gap still wants explaining, but a transcription with a perfect record and one with two known errors are not equal witnesses. Still not settled, and only the page settles it. Senj births are filmed 1878–1894, so ahnentafel 3's entry is reachable. One opening decides whether Hedviga's mother was born in October or November. --- 20 September 2026 — this row's «ahnentafel 3 and 6 are unread» is now stale in both halves, for different reasons. Ahnentafel 6 is read. `josip-anton-papic` carries his baptism of 1 November 1857 and his marriage of 3 May 1885, plus a grave and three derived credits. Those rows existed all along but were filed under `josip-anton-JOSO-papic`, a slug that names nobody, so every tool skipped them — see row 16 and [notes/eight-readings-that-named-nobody.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/eight-readings-that-named-nobody.md). Repairing the slug is what made him read. Ahnentafel 3 is found and deliberately not recorded. Her baptism is `1:1:X9D5-SHTB`, Senj, 17 November 1886 — and the index gives her birth as 9 November where the tree says 9 October. [notes/senj-1886-milka-baptism-index.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/senj-1886-milka-baptism-index.md) sets out why her row stays out of `readings.psv` until somebody opens the page: writing it in would either assert an index over an image or bury a conflict in a citation. Senj births are filmed 1878–1894, so one opening settles it. What did come out of the parish today is ahnentafel 7, at Krivi Put rather than Senj — see row 31.

All three read, 15 September 2026. Ahnentafel 27 (9 Feb 1886), 8 (26 Jul 1907) and 26 (14 Nov 1897) are each at the page, and every date matches the tree to the day. What they add. Ahn 26 is *«finan.… u miru»* — a retired finance man; ahn 8 is a *nadničar*, a day labourer. Neither had an occupation here before, and the two Senj households sit either side of a line. Karlobag is recorded as the origin of *both* ahn 26 and ahn 27, confirming the birthplaces the tree carries and putting the whole couple in Karlobag before Senj. And ahn 8's entry touches ahnentafel 9 for the first time — *«Udovac iza pok. Tereze»*. The finding worth more than the entries: the age column is hearsay, and the clerk says so. Ahn 26's age is written *«navodno 93 god.»* — *reportedly* 93 — against a true 91. Ahn 27's reads 78 against 75. Ahn 8's reads 69 and is exactly right. An age in this book is what the informant said, never a document, and it must never be used to date a birth where a baptism exists. Two indexing traps recorded. Ahn 27 is filed under Pilipić, her married name — a search for *Uroda* finds nothing. Ahn 26 is filed under Pilepić, a different spelling from his own wife's entry in the same office eleven years earlier.

100 people, 58 reachable, carrying ahnentafel 7. Only one Krivi Put page has ever been opened here (the 1859 baptisms, recorded *partial*). The catalogue range was verified on 14 September, so the book is known to cover the window. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Krivi Put IS indexed — and the first probe said it was not. Searched under Prpić, this parish returns gravestones and nothing else. Searched under Tomljanović, it returns 30 indexed baptisms, 1850–1870, in *Hungary, Croatia-Slavonia, Births*. A single surname can miss an entire collection, and this row's own surname is the one that misses it. And there is a death index here, the only parish register death index in the table: *Hungary, Croatia-Slavonia, Deaths*, 1891–1924. Too late for ahnentafel 7's baptism, and not too late for her death. --- 20 September 2026 — searched by parents, and the household came out. Ahnentafel 14 × 15, Joannes Perpić and Lucia, at Krivi Put. The index writes her «Lucia Perpić» — independent corroboration of question 5c, which recovered her surname off the 1885 marriage at nine times magnification as *«Luca rodjena Pèr-pić»* and concluded she is a Prpić who married a Prpić. The index agrees, from a different document. Three children the tree does not have — Georgius 22 April 1868, Joanna 6 January 1874, Josephus 7 July 1876 — all recorded in `found.psv` as candidates, because Prpić is the dominant surname of this parish and the index holds several *Perpić × Perpić* couples. What supports them is that the same couple's Stephanus of 1861 is in the tree. And he brought a second date conflict of exactly the shape Milka's has. The tree gives Stephanus 13 OCTOBER 1861; the index gives 13 December 1861, baptised the 16th. The day survives, the month does not, and the tree's month is the earlier one — as it was for ahnentafel 3. Four other dates from the same sweep agree exactly. A mechanism, offered as a hypothesis: these registers write the month as a Roman numeral, and both errors are the same error — *XI → X* and *XII → X*, the trailing strokes lost. That would preserve the day, corrupt only the month, and could only ever make the month earlier. It predicts a direction, so it is testable: the tree's month should never be *later*, and the disagreements should cluster in October, November and December. Six comparisons is a shape, not a test. [notes/the-month-is-wrong-and-the-day-is-right.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-month-is-wrong-and-the-day-is-right.md) And the Roman-numeral hypothesis did not survive the test — see row 29. Seventeen image-read dates agree with the tree exactly; both month conflicts are against the INDEX, not against a page. Stephanus's 13 October 1861 is more likely right than the index's 13 December. --- 20 September 2026 — ahnentafel 7's own entry was in the result set that produced her brothers, and it was passed over. `found.psv` has carried Georgius 1868, Joanna 1874 and Josephus 1876 as candidate siblings since this morning, and Stephanus 1861 is quoted in the month note. Antonija «Tonka» Perpić had a gravestone, four derived credits, and no entry of her own. Birth 12 May 1859 · Baptism 15 May 1859 · Krivi Put Father «Joannis Perpić» Mother «Luciæ Perpić» · ark 1:1:XQM9-6NQ4 It agrees with the tree to the day, and the three-day gap to baptism is ordinary here — which is a third data point saying this index is not careless about this family, against the two entries that disagree with the tree by a month. It gives ahnentafel 15 her first document of any kind. She had none. It does not give her a maiden name: the Krivi Put index writes every mother under the father's surname, so «Luciæ Perpić» is a convention. Two more candidate elder siblings into `found.psv` — Martha 1853 and Toma 1856 — both flagged CANDIDATE, because at least two Joannes Perpić households sit at Krivi Put in this decade and one of them is married to a Francisca. [notes/the-household-was-swept-and-she-was-left-out.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-household-was-swept-and-she-was-left-out.md)

132 people, 120 reachable — the highest count of any book here — and not one direct ancestor. That combination is exactly why it keeps being deferred, and exactly why it should be done deliberately rather than by accident: it is breadth, not depth, and it would settle a great many collateral lines at once. It is also the earliest register in the estate — the 1650 book was opened once and its earliest leaves are in Croatian, not Latin, which is a finding in its own right. Do it when the direct line is exhausted, or when a wet afternoon wants volume. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Novi Vinodolski baptisms are indexed, 1826–1899, in *Hungary, Croatia-Slavonia, Births*. This row has been deferred for months as *«the largest reach in the archive»* — 132 people, 120 reachable — on the assumption that it meant opening the earliest register in the estate. For the nineteenth century it does not: it is a search box. The 1650–1825 stretch is still a walk, and that is where the row's «earliest register» value lies. --- 20 September 2026 — this book is indexed far wider than the coverage table said, and it is Klenovica's parish. `indexed.psv` recorded Novi Vinodolski as *baptism, probed 1680s 1850s*. A place-only census, naming no surname: | births | entries | | births | entries | |---|---|---|---|---| | 1650–1749 | 300 of 300 | | 1830–1859 | 300 of 300 | | 1750–1799 | 300 of 300 | | 1860–1879 | 300, baptisms and burials | | 1800–1829 | 300 of 300 | | 1880–1899 | 300, baptisms and burials | Dense in every window from 1651 to 1899, and burials are indexed from 1860 — about a quarter of the rows in the last two windows and none before. A book recorded as sampled in two decades is searchable across two and a half centuries. And it is the Kalanj parish. `parishes.psv` gives Novi Vinodolski as *novi vinodolski, klenovica, povile*, and 101 of this archive's people are at Klenovica. This row says «no ancestor in it» — true of the direct line, and it undersells what is here. It has already produced one. `corrections.json` records four Kalanj baptisms «sitting unattached in the family's own photo library». The index holds five: a Rozalia Kalanj, baptised 5 August 1893, ark `1:1:XQMQ-QY3Q`, same parents — and `rozalia-kalanj` is in the tree with no reading of any kind, because hers is the one scan nobody had. The four that were in the library got read; the fifth was never looked for. Filed INDEX ONLY and flagged: the tree gives her birth as 4 July and the index her baptism as 5 August, a month where her siblings' entries put days. [notes/novi-vinodolski-is-indexed-to-1899.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/novi-vinodolski-is-indexed-to-1899.md) And then the opening closed, the same afternoon. A densely indexed parish holding 101 of this archive's people looked like the best lead of the day. It is not. The Kalanj surname appears FIVE times in the whole Novi Vinodolski index — eight spellings tried, as surname and as father's surname — and all five are the one household of Martin Kalanj and Marija Peričić, 1893 to 1900. Nothing earlier, nothing else. It is not surname-blindness, and it was tested the way that fault is tested. *Father Josephus × mother Rosalia, 1855–1870* — the household whose baptisms `corrections.json` quotes off the page — returns 0. *Josephus, 1824–1828* returns 24 dense Novi Vinodolski baptisms: Sokolić, Krišković, Maričić, Petrinović, Karlović, Potočnjak, Piskulić, Radetić and Peričić. Peričić is in the index and Kalanj is not, same parish, same decade. The index for this parish does not hold this family before 1893. And the village is invisible. «Klenovica» as a place returns Cres, Hreljin, Lopar — other parishes entirely; *Povile* the same. Every row here is filed «Novi Vinodolski», 200 of 200 in the 1800–1860 sample. Search the parish, never the village. So this row stays a film read — baptisms 1650–1900, marriages 1674–1859, deaths 1704–1899 per `parishes.psv` — in a parish whose index is dense for everybody except us. The five indexed entries are the end of what the search box can do here, not the beginning.

The two Blažević ancestors without a document, ahn 16 (Anton) and ahn 32 (Ilija, b. 1790), are both placed at Smokvica Krmpotska. Krmpote-Vodice births score 15 reachable and Ledenice 15 more. The trap is recorded and must be respected: Krmpote's catalogue label says 1815 and the film actually begins in 1879 — verified by opening it — so Ilija of 1790 is *not* in it. Ledenice (1734–1894) is the book that could hold him, and Krmpote is administered from Ledenice to this day. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Krmpote is indexed 1880–1894 — and the index confirms the trap this row records. The catalogue label says the film begins in 1815; this row already established by opening it that it actually begins in 1879. The index agrees with the film and not the catalogue. Ahnentafel 32, Ilija, born 1790, is not reachable by index or by film, and Ledenice returns nothing indexed at all. --- 20 September 2026 — re-tested with the date-window method, and the answer is firm. The first coverage pass probed broadly and I had to correct it within the hour, because a broad query is ranked and capped and whole collections hide below the cut. So Ledenice was left in `indexed.psv` marked *«broadly probed only — re-test by window»*. Re-tested, under Blažević and Tomljanović, across 1730–1800, 1800–1850 and 1850–1900: Ledenice returns nothing at all. Krmpote returns only the 1850s window, matching the 1880–1894 already observed and the film that begins in 1879 rather than the catalogue's 1815. Smokvica Krmpotska returns nothing under Blažević. So ahnentafel 16 and 32 cannot be reached from a search box, and this row is a film read — now measured twice by two methods rather than assumed. And the shape of the whole thing is worth naming. The indexing in this archive's country is a Karlobag–Otočac–Senj phenomenon. Those three are indexed deeply, by several collections each, back to the seventeenth century. Smokvica, Krmpote before 1880, Ledenice, Selce, Brinje, Bribir, Sinac and Gospić are not indexed at all. Hedviga's mother's side is searchable and her father's side is not — which is why every finding of the last two days has been a Papić, a Pilipić, an Uroda or a Žubrinić, and not one has been a Blažević. --- And the one Blažević-side window that IS indexed was searched — Krmpote 1880–1894 — with two results, one of them about question 1b. The index gives Marijan Blažević the wrong father. It holds `1:1:XQMS-HRQW` — *«Marijan Blažević, baptism 1895 Krmpote-Vodice, birth Dec 1895, Ivko Blažević × Tereza Zubrinić»*. This archive read the same entry off the image on 9 September: p.168 no.26, «1895, travnja 12 / 16» — *travanj is April* — parents *«Jurko Blažević i Tereza rođ. Žubrinić, selj.»*, Smokvica kbr. 114. It is one entry, not two — two entries would need two women called Tereza Žubrinić married to two Blažević men in one small parish, each naming a son Marijan in 1895. And the entry shows how the name went wrong. The GODFATHER is «Ive Blažević, sin Tomin». Both names are on the line, and the index has put Ive in the father's slot. That is a mechanism rather than a guess, and it leaves the image read intact — with the house number and the occupation the index does not carry. It changes nothing in question 1b, which is the point of checking. But anybody searching the index rather than the film will be told Marijan's father was an Ivan, and will not find Juraj at all. [notes/krmpote-1895-the-index-took-the-godfather.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/krmpote-1895-the-index-took-the-godfather.md) And the negative: Ljubomir (1892) and Anka (1893) are both inside the indexed window and neither baptism is there. The only Ljubomir entry anywhere is his 1920 marriage, the only Anka entry her 1919 marriage. The couple's two eldest are not in the Krmpote baptism index even where it covers their years — consistent with question 13, which argues from the other direction that Ljubomir was baptised at Rijeka. --- 20 September 2026 — Ledenice is not a zero either, and it is still a film read. Both halves matter. `indexed.psv` said NOTHING for Ledenice, tested under Blažević and Tomljanović. A surname-free place probe finds four entries across two centuries: 1748, 1749, 1845 and 1888. That is not a register, and this row's plan does not change — Smokvica is still zero and Krmpote before 1880 is still zero, so the Blažević side remains unreachable from a search box. But the 1888 one paid. «Marija Milka, 16 May 1888, Ledenice, parents Luka Kalanj and Kata», ark `1:1:XQM9-8KKX` — and `marija-milka-kalanj` is in the tree, carrying the bare year *1888* and a citation to the register image. The index supplies the day the tree does not have. Filed `INDEX ONLY`; the index gives her mother the father's surname where the tree has *Kata Miletić*, which is an indexing convention rather than a conflict. And a caution this row should carry. The same place-only probe returned zero for Krmpote in every window — while this archive found an indexed Krmpote-Vodice baptism of 1895 by surname on the same day. Neither probe is authoritative alone. A zero from either means «not found by this instrument».

Of the 34 direct ancestors without a document, 21 have no place recorded — not a village, not a parish, in several cases not a date either: ahn 22, 41, 64, 80, 84, 85, 86, 87, 97, 100, 101, 102, 103, 109, 128, 210, 211, 213, 427, 428, 429. No amount of film-reading reaches them, because nothing says which film to open. This row is the precondition for half the rows above it. The method that works is the one that found ahnentafel 55's father: a person with no record of their own is named in their child's. Every one of these twenty-one has a child in the tree; the child's baptism or marriage entry names the parent and usually the village. So the job is to work down from the children, not up from the parents. --- Two came off the list on 15 September without any film being opened, because they were never placeless. The list was built from the wall grading, and the wall grading called a household «No place recorded» whenever no parish in `parishes.psv` claimed its village. That is not the same thing. The tree gives ahnentafel 426 as born at Smiljan, Općina Gospić and ahnentafel 222 at Montani Stanište — both written plainly, both printed for weeks under «*No place is recorded for this household anywhere on the line*». `walls.py` now separates the two, with a new grade `ungazetteered` — *«The tree does record where this household was … but no parish in this archive's table claims that village … This is not a wall in the record. It is a gap in the gazetteer.»* Nineteen of the twenty-one are genuinely placeless. Two are one line of `parishes.psv` away, and that line is row 56. A third came off earlier the same day: ahnentafel 220, placed at Karlobag by his own death entry. So the twenty-one is eighteen, and the method the row proposed — *work down from the children* — has not been started yet. --- Three more came off the list on 15 September, and again no film was opened. The wall grading looked at one field on two people — the person's own `bornPlace`, then the ahnentafel child's — and called everything else *«No place is recorded for this household anywhere on the line»*. That sentence was false three more times. * Ahnentafel 102, Joanni Brozović — his daughter Margarita is recorded as dying at Selce. A death place is a place. * Ahnentafel 103, Catharina, his wife — the same daughter, the same parish. * Ahnentafel 427, Mara Pekass — her husband is recorded as born at Smiljan, and the grading never once looked at a spouse, though it had been computing one all along. A husband's parish is his wife's. The fallback now runs through the household, nearest first — own death, child's birth, child's death, spouse's birth, spouse's death — and each wall says which relation supplied it: *«This household is put at Selce on the strength of their child's death — the tree's own words about the tree's own people, not a document.»* That distinction is the whole point. Nothing here was read off a page, and a wall that pretended otherwise would be worse than one that said nothing. Four are genuinely placeless and always were: ahnentafel 84 and 86, Vitus and Michael Sekula; 128, Mateša Blažević; 428, Angelo Bacchi. Not a village, not a death place, not a spouse, not a child with either. For these four the row's original method — work down from the children's baptisms — is still the only way in, and it still needs a register. --- 17 September 2026 — and the four «genuinely placeless» were three fields short, not a register short. The fallback ran own death → child's birth → child's death → spouse's birth → spouse's death and stopped. It stopped one generation too early, and «*no place is recorded for this household anywhere on the line*» was false a third time — over the very sentence that says *on the line*. * Ahnentafel 128, Mateša Blažević — his grandson Ilija is born at Smokvica Krmpotska, written plainly in the tree. * Ahnentafel 84, 85, 86 and 87 — Vitus and Michael Sekula and their two wives called Maria — all four are grandparents of ahnentafel 21, born at Otočac. One person's birthplace places four walls. * Ahnentafel 428, Angelo Bacchi, and 429, Catharina — their granddaughter Oliva Andreana Bacchi dies at Karlobag. The grandchild goes last in the chain because it is the weakest, and each wall says so in the same words the others use: *«This household is put at Karlobag on the strength of their grandchild's death — the tree's own words about the tree's own people, not a document.»* Nothing was read off a page. Unplaced went from four to none — and two of them are now reachable books. Angelo Bacchi's baptism of about 1752 and his marriage to Catharina of about 1777 both fall inside Karlobag, which is filmed 1691–1861 and indexed — that is row 54's finding. Catharina has no surname at all, and a marriage entry is what supplies one. Two walls moved from «nothing is known» to «a book this archive already knows how to read». The rest are honest silences: 84, 85, 86, 87 want Otočac before 1834, and 128 wants Krmpote before 1879. Neither film goes back that far. --- 21 September 2026 — the one wall the tool could not grade at all has been measured, and the answer is «Karlobag, evidenced» rather than «Karlobag, assumed». Of thirty-one walls, exactly one is graded `unchecked — ranges not checked`: ahnentafel 222, Pauli Tomljanović, placed at Montani Stanište, which is a herders' settlement and not a parish, and which the FamilySearch catalogue does not hold under either word. Its wall said the Tomljanovići here «are Karlobag, which makes Karlobag the first place to look and not an answer». It is a better answer now. The Karlobag record index holds 120 Tomljanović rows between 1721 and 1820 — *Tomlenovicha, Tomlinovich, Tomlienevics, Tomljenovich, Tomlenovich* — so the surname is native to that parish and not merely this archive's habit. Combined with ahnentafel 111's six children baptised at Karlobag 1787–1801 and her death there in 1809, Karlobag is where this household's paper is. A candidate for ahnentafel 222, offered as one. A Paulus Tomljanović × Thoma fathers at Karlobag in those years — Josephus, July 1784 and Joanna, June 1787 — and ahnentafel 222 is the father of a Lucia bearing children from 1787, so born about 1765. Both children are in `found.psv`. And the reason it is only a candidate is on the same page. A second Paulus Tomljanović is fathering at Karlobag in the same year by a Catharina — *Matthæa, 7 October 1787*. Two Pauluses, two wives, one parish, one year. Ahnentafel 111's mother is unknown, so nothing ties her to the Thoma rather than the Catharina. The third entry is in `found.psv` specifically to keep them apart, because the next reader will meet one Paulus and assume there is only one. Lucia's own baptism is not in the index under any spelling — every indexed Lucia Tomljanović at Karlobag is 1818 or later. Her generation is below the index's reach for this surname, which is the same shape as the Kalanj at Novi Vinodolski. The mapping is deliberately not asserted. `parishes.psv` still gives Montani Stanište as `unknown`, with the measurement recorded beside it: a surname being native to a parish is not the catalogue placing a *stanište* in it, and this archive does not close a gazetteer gap with an inference. --- 28 September 2026 — the target list this row said was the method, actually built. For each of the twenty-one, the tree's own data was asked: *which child of theirs has a record with a place on it?* The answer is the parish whose film would name the parent. | Ahn | Who | The child whose baptism names them | Year | Parish | |---|---|---|---|---| | 109 | Maria Uroda | Joannis Georgius (Georgio) Uroda | 1777 | Karlobag | | 210 | Vicentius (Vicko) Luckinić | Catherina (Catta) Luckinić | 1727 | Karlobag | | 211 | Maria Luckinić | Catherina (Catta) Luckinić | 1727 | Karlobag | | 213 | Antonia Pekass (Pekkas) | Vicentius Fabiano Smojver Vukić | 1754 | Karlobag | | 97 | Ursula | Joannis Antić | 1765 | Selce | | 100 | Mathia Jelličić | Georgi Juraj Jelličić | 1767 | Selce | | 101 | Margarita | Georgi Juraj Jelličić | 1767 | Selce | | 102 | Joanni Brozović | Margarita Brozović | 1784 | Selce | | 103 | Catharina | Margarita Brozović | 1784 | Selce | | 41 | Marija Findrić | Mihovil ŽUBRINIĆ | 1802 | Senj | | 80 | Žubrinić | Toma Žubrinić | 1768 | Senj | | 64 | Grgica Blažević | Ive Blazevic | ? | Sv. Jakov Krmpote | | 22 | Shambul Marić | Mathias Marić | ? | — | | 84 | Vitus Sekula | Susanna Sekula | 1823 | — | | 85 | Maria | Susanna Sekula | 1823 | — | | 86 | Michael Michaelis Sekula | Mate Matthew Matthias Mattheu Sekula | 1822 | — | | 87 | Maria ? | Mate Matthew Matthias Mattheu Sekula | 1822 | — | | 128 | Mateša Blažević | Grgica Blažević | 1770 | — | | 427 | Mara (Maria) Pekass | Antonia Pekass (Pekkas) | ? | — | | 428 | Angelo Bacchi | Caetano Bacchi | 1779 | — | | 429 | Catharina | Caetano Bacchi | 1779 | — | And it collapses into three films. * KARLOBAG — four of them (ahn 109, 210, 211, 213), children baptised 1727–1777. That is row 54's film exactly: Karlobag baptisms 1691–1861, DGS 5481648, which that row already wants for four *other* walls. Eight walls, one film, one sitting. * SELCE — five (ahn 97, 100, 101, 102, 103), children baptised 1765–1784. The Selce deaths are already being read for row 25; the baptisms are not. * OTOČAC and SENJ — one each (ahn 80 at Šumećica 2 in 1768, ahn 41 at Senj in 1802), and ahn 64 at Smokvica Krmpotska, which is Sv. Jakov Krmpote — the parish `/blazevic/` names and nobody has opened. Nine of the twenty-one are not reachable this way and should stop being counted as if they were: ahn 22, 84, 85, 86, 87, 128, 427, 428 and 429. Their children carry no place either — Sekula, Pekass, Bacchi and Marić are placeless two generations deep, so the method that works for the other twelve does not touch them. That is a different problem and it needs a different row. This list needs no login to make and every film on it needs one to read. It is built and waiting.

Founded on a mistake, and kept here rather than deleted — the mistake was in the scope, not in the work. This row asked which register serves Gologorica, where `targets.json` puts 23 Šestani. Two sittings went into it. Nobody asked the prior question: why is a Šestan row in a Blažević archive? It is not one. `tools/ancestry.py` carries eight surnames and `sestan` is among them, so 55 Šestani are published — but: * Twenty-five of the fifty-five have a DEFRANCESCHI mother — Joanna (10 children), Sancta-Alexandra (9), Rosa (6). * No Šestan connects to Hedviga through any published person at all. In the full 15,643-person tree the nearest is ten steps away: `Hedviga → married Ivan Anton Defranceski → up five Defranceschi generations → (unnamed) → «Grah's for Investigation (not real)» → «Gologorica Grah» → Josephus Grah 1864 → married Rosa Sestan 1872` * Two of those links are sorting buckets, and one of them — *«Grah's for Investigation (not real)»* — this archive's own `buckets.py` already classifies as not a person. So they are reached through Hedviga's HUSBAND's family, and then only through a node the archive itself says is not real. The Defranceschi archive's own queue had the corroboration waiting: Forebears gives Gračišće's surnames as *Jugovac 489, Šestan 316, Pauro 234* — Šestan is the second commonest surname in that village. Handed over 20 September 2026 to `TheDefranceski`, commit `1cbcea8`, appended to their `notes/QUEUE.md`: that Gologorica is not a catalogued parish and the zero was tested against a nonsense control; that Pazin's book reaches 1582 (`koha:725172`) with the film numbers; that the Šestan baptisms are indexed under PISINO and invisible under Pazin; that nothing yet places the 23 Gologorica Šestani in that book, so the film is not yet worth a sitting; and two free people — a death and a married name for Elisabetha Sestan, and «Gologorica Grah», a village name sitting in a given-name field. One thing is left open and is not this archive's to decide: whether `sestan` should stay in the eight-surname scope list. Taking it out would unpublish 55 people. It is recorded in [the sister-archive notes](https://github.com/daviddef/TheBlazevic/blob/main/notes/for-the-sister-archives.md) rather than acted on. --- What the two sittings actually found, kept because it is true and was paid for: `parishes.psv` carries Gologorica with *unknown* in all three range columns and the note «Istria — likely a different collection entirely». `targets.json` puts 23 people behind it. It is the only parish in the table that has never been resolved at all. One catalogue search decides whether it is reachable or whether it joins the unreachable list honestly. --- Half answered, 15 September 2026, and the half that is answered changes its priority. Every one of the 24 people behind Gologorica is a Šestan, and not one of them is a direct ancestor. The parish stands entirely on a married-in line. That does not make it worthless — it makes it the last of the twenty-one parishes to spend a day on, not the outstanding hole the row implies. The catalogue question is still open and could not be answered from here. The catalogue's own search service refuses every parameter this archive knows how to send (`q.text`, `q.placeText`, `q.anyPlace` all return 400; the UI's own call to `/service/search/catalog/v3/search` returns 400 too and the page prints *«Something Went Wrong»*). The place search wants a place id, not a name, and getting one needs the typeahead the interface uses. So the row stays open with a smaller question: not *«is Gologorica reachable»* but *«which collection holds Istrian parish registers, and is it this one»* — and it can wait behind every row that touches the direct line. --- 20 September 2026 — the one catalogue search, done. The answer is neither «reachable» nor «unreachable». Gologorica is not a catalogued parish. Searched under four forms — *Gologorica*, *Gollogorizza*, *Gologorizza*, and the modern municipality *Cerovlje* — all empty. And the zero was tested rather than assumed, because this catalogue pads every result with a constant filler list: a Staffordshire map, a Wagler family history, a Hungarian church in Hamilton. Thirty rows come back whatever you ask. | query | result | |---|---| | `Karlobag` | koha:655222 «Matična knjiga, 1691-1861», then filler | | `Gologorica` | filler only | | `zzzzqqqnotaplace` | filler only — byte-identical to Gologorica | But the district town is catalogued, and its book is extraordinary. `koha:725172` — «Matična knjiga, 1582-1882», filmed from the Hrvatski državni arhiv: births 1582–1882, marriages 1597–1852, deaths 1655–1847, with `koha:725059` adding births 1656–1822. Births from 1582. The earliest this archive has reached until today is Novi Vinodolski, 1650 — which row 32 calls *«the earliest register in the estate»*. Pazin is sixty-eight years older. And the catalogue note is the sovereignty problem in one sentence: *«for Mitterburg, Küstenland, Austria; later Pisino, Istria, Italy; now Pazin, Croatia»* — one parish filed under three countries and three names. This archive met the same thing from the other end four days ago, when one man gave *«Zara, Austria»* in 1919 and *«Italy»* in 1924 because Rapallo happened in between. A searcher who knows only the Croatian name finds a third of this parish. `parishes.psv` now carries Pazin in place of the unresolved Gologorica row. What is NOT established is that the Šestani were written into it — a village ten kilometres from a district town may have had its own chapel or belonged to a nearer parish that is itself uncatalogued. [notes/gologorica-is-not-a-parish-but-pazin-is.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/gologorica-is-not-a-parish-but-pazin-is.md) And the index was then asked what the catalogue could not answer — is Pazin's book the Šestani's book? There is no evidence that it is. The Šestani ARE indexed at Pazin, in the eighteenth century — searched as Pisino: births 1795, 1804, 1807, 1826, 1827, 1834, deaths back to *Joannes Sestan, born 1659, died 23 April 1709*, marriages from 1787. Searched as «Pazin» those births do not appear at all — the collection is filed under the Austrian name. That is the three-names problem costing a searcher two thirds of a parish, shown on this family's own surname. But not one of this archive's Šestani is among them. `targets.json` puts 23 Šestani at Gologorica, born 1843–1877. In that window the index returns Šestani at Poreč, Rovinj, Pola, Ližnjan and Vrsar — and none at Gologorica, none at Pazin. Šestan is an Istrian surname spread across the peninsula: the Kalanj lesson and the Perpić lesson a third time. So: the Pazin film is available and is not yet worth a sitting. Writing that down is the point — without it the next reader sees *1582* and spends a day on a book that may have nothing to do with this family. One record does put a Šestan at Gologorica, and it is American: *Elisabetta Sestan Rabar*, born 11 July 1912 at *«Gologorica I\*, Yugoslavia»*, died Staten Island 2005, parents *Natale Sestan × Antonia Ancich*. The village is real and people were born there. It is the parish register that is missing, not the place. And a loose thread kept rather than tidied: *«23 March 1842, Otočac — Johanna Sestan, of Philippus Sestan × Martha»*. A Šestan baptised in Lika, three hundred kilometres from Istria, in this archive's own heartland. Nothing is built on it. --- And the last question was the one nobody had asked: whose family is this? Twenty-five of the fifty-five Šestani in this tree have a DEFRANCESCHI mother. Three women — Joanna (10 children), Sancta-Alexandra (9), Rosa (6). The Šestani are here because three Defranceschi women married Šestan men. They are not a Blažević family and not a Lika one, and the 23 at Gologorica are this archive's people only by marriage. Which makes this the Defranceschi archive's row. That archive works Gračišće and the county of Pazin — the same district as Gologorica, the same `koha:725172` from 1582, the same Mitterburg / Pisino / Pazin problem. Everything found here today is more use to it than to this one: that the village is not a catalogued parish, that the district book reaches 1582, that the births are filed under the Austrian name and invisible under the Croatian one, and that Šestan is a peninsula-wide surname. This archive does not hand work away silently, so the row says so rather than being struck. The owner is now *«this archive, and then Defranceschi»*. One person gained a death and a husband on the way past — and it is a rediscovery, which is worth admitting. The American record offered above as *«one record does put a Šestan at Gologorica»* is somebody this tree already holds: `elisabetha-sestan-3`, Elisabetha Sestan, born 11 July 1912, of *Natalis Joannes Sestan × Antonia Ancic*. She was presented as a stranger. She is not. What the record adds is the end of her life, not the beginning — *«Elisabetta Sestan RABAR, died 20 April 2005, Staten Island»* — and the tree gives her no death and no husband. Married name Rabar, died in New York at ninety-two, recorded in `readings.psv` as an index and not a certificate.

`consistency.json` holds 71 contradictions and the largest class is child-after-parent-death, 25 of them. Each is either a wrong date, a wrong link, or two people merged into one — and this archive has already found all three kinds. They are cheap to triage against the register readings now on disk, and every one resolved either corrects the tree or produces a `corrections` row. --- Triaged 15 September 2026, and the triage found something the four rows did not ask about. Every contradiction was supposed to be answered on the person's own page, and `check_site.py` has a gate that says so. It read `if i.get("slug") and …` — so it quietly excused every contradiction about a person with no page. That is 47 of the 71 rows and 24 people, none of them answered, and the check printed ok the whole time. It is the same blindness for the third time in one day. Six Karlobag wives were declared outside the tree because they had no page. Hedviga's husband's gravestone was filed as a stranger's because he has no page. Now the contradiction gate. An unpublished person is invisible, in three different tools, and each time it was found by accident. The gate now counts and prints them rather than excusing them — failing only on the slugged ones, the same bargain `checkcovers` struck so a hole can be closed without the archive going red first. Row 55 carries the backlog. And of the 24 rows that WERE answered, six belong to one bad date. Ahnentafel 9, Tereza Pećanić, is entered as «*died BEF 1850*» and then marries in November 1858 and bears children until 1884 — five *child-after-parent-death* rows and one *married-after-death*, all from a single hedged date that cannot be right. It is annotated, and it is still sitting in the data generating contradictions.

The second and fourth classes in `consistency.json`: parent-too-young (16) and parent-under-seventeen (8). In a district where the register routinely records a bride's age, these are testable rather than arguable — and the commonest cause in this tree so far has been a grandparent and a parent sharing a name, which is precisely the trap row 1 documented for the Otočac Michaels. --- Triaged 15 September 2026, and the triage found something the four rows did not ask about. Every contradiction was supposed to be answered on the person's own page, and `check_site.py` has a gate that says so. It read `if i.get("slug") and …` — so it quietly excused every contradiction about a person with no page. That is 47 of the 71 rows and 24 people, none of them answered, and the check printed ok the whole time. It is the same blindness for the third time in one day. Six Karlobag wives were declared outside the tree because they had no page. Hedviga's husband's gravestone was filed as a stranger's because he has no page. Now the contradiction gate. An unpublished person is invisible, in three different tools, and each time it was found by accident. The gate now counts and prints them rather than excusing them — failing only on the slugged ones, the same bargain `checkcovers` struck so a hole can be closed without the archive going red first. Row 55 carries the backlog. And of the 24 rows that WERE answered, six belong to one bad date. Ahnentafel 9, Tereza Pećanić, is entered as «*died BEF 1850*» and then marries in November 1858 and bears children until 1884 — five *child-after-parent-death* rows and one *married-after-death*, all from a single hedged date that cannot be right. It is annotated, and it is still sitting in the data generating contradictions.

`consistency.json`'s child-before-parent-birth, 10 rows — plus the singletons died-before-born, impossible-generation-gap and married-before-birth. These cannot be softened by interpretation the way an age gap can; each is a definite error in the data and should be resolved or struck. --- Triaged 15 September 2026, and the triage found something the four rows did not ask about. Every contradiction was supposed to be answered on the person's own page, and `check_site.py` has a gate that says so. It read `if i.get("slug") and …` — so it quietly excused every contradiction about a person with no page. That is 47 of the 71 rows and 24 people, none of them answered, and the check printed ok the whole time. It is the same blindness for the third time in one day. Six Karlobag wives were declared outside the tree because they had no page. Hedviga's husband's gravestone was filed as a stranger's because he has no page. Now the contradiction gate. An unpublished person is invisible, in three different tools, and each time it was found by accident. The gate now counts and prints them rather than excusing them — failing only on the slugged ones, the same bargain `checkcovers` struck so a hole can be closed without the archive going red first. Row 55 carries the backlog. And of the 24 rows that WERE answered, six belong to one bad date. Ahnentafel 9, Tereza Pećanić, is entered as «*died BEF 1850*» and then marries in November 1858 and bears children until 1884 — five *child-after-parent-death* rows and one *married-after-death*, all from a single hedged date that cannot be right. It is annotated, and it is still sitting in the data generating contradictions.

`impossible.json` carries 45 rows in four causes: record 27, generation 9, hedged 5, namesake 4. The *namesake* four are the same class of problem as the seven Michael Žubrinići, and the method that separated them — the house number in the Conditio column — now exists. Worth re-running the four against what row 1 established before treating them as insoluble. --- Triaged 15 September 2026, and the triage found something the four rows did not ask about. Every contradiction was supposed to be answered on the person's own page, and `check_site.py` has a gate that says so. It read `if i.get("slug") and …` — so it quietly excused every contradiction about a person with no page. That is 47 of the 71 rows and 24 people, none of them answered, and the check printed ok the whole time. It is the same blindness for the third time in one day. Six Karlobag wives were declared outside the tree because they had no page. Hedviga's husband's gravestone was filed as a stranger's because he has no page. Now the contradiction gate. An unpublished person is invisible, in three different tools, and each time it was found by accident. The gate now counts and prints them rather than excusing them — failing only on the slugged ones, the same bargain `checkcovers` struck so a hole can be closed without the archive going red first. Row 55 carries the backlog. And of the 24 rows that WERE answered, six belong to one bad date. Ahnentafel 9, Tereza Pećanić, is entered as «*died BEF 1850*» and then marries in November 1858 and bears children until 1884 — five *child-after-parent-death* rows and one *married-after-death*, all from a single hedged date that cannot be right. It is annotated, and it is still sitting in the data generating contradictions.

`duplicates.json`: 51 groups, of which 23 are live and 28 already rejected. A live duplicate group is two tree records that may be one person — which is the same question as the two Georgius Urode and the seven Michaels, and it now has a settled method. 28 have already been rejected, so the remaining 23 are the hard ones and deserve the register rather than judgement. --- Worked 15 September 2026. Six of the twenty-three are settled, and the tool had only half a test. `duplicates.py` had `parents_conflict` — a veto that rejects a group when two records name parents sharing no name. It had no confirmation: two records naming *the same* parents stayed «live» exactly like two naming nobody. So a new tier, `same day and same parents`, and five groups fall into it — same birth day, same village, same father, same mother, nothing left for a second child to be: | Who | Born | Parents | |---|---|---| | Catharina Žubrinić | 20 Nov 1840, Otočac | Mate × Anna Mudrovčić | | Josephus Žubrinić | 16 Mar 1845, Otočac | Michael × Maria Marković | | Francisca Žubrinić | 6 Mar 1846, Otočac | Thomas × Joanna Ratković | | Nicolaus Žubrinić | 3 Aug 1847 | Thomas × Joanna Ratković | | Katalyn / Kate Zubrinich | 26 Jul 1886, Port Pirie | John × Rose Gormley | And Danijel Kalanj, 2 November 1829, Klenovica is the plainest of all without qualifying: one copy carries his parents, the other his wife and child, and both carry the same death date, 26 June 1876. One man, entered twice, split down the middle. The sixth is not a duplicate at all, and finding that out is the point. «*Alexander Žubrinić, 14 March 1839, Otočac*» has the same day and the same village and a Michael for a father in both — and the mothers are Magdalena Orešković and Magdalena Mande Draženović. Two boys, born the same day, to two different Michaels. It is row 1's seven Michaels in miniature, and the first draft of the new tier declared it a duplicate, because `parents()` folds a parent to their given names only and both mothers reduce to *magdalena*. `parent_surnames()` now supplies the half of the name that was being thrown away. And the house number cannot rescue it. Alexander is 1839, and this archive established that the Otočac baptisms **stop writing the household into *Conditio* from about 1836 to 1845. The one instrument that separates Michaels does not reach the years where they most need separating. The 1834–1858 deaths do carry it — that is row 28's finding — so the way in is a death, not a baptism. Seventeen groups remain**: eight *same year and place*, eight *same year only*, one *same day*. They are the hard ones, as the row said. --- A second sitting, and the tool had a second blind spot — this one in its keys. Every group above is found by keying on a surname and a given name, so two records can only meet if somebody spelled them alike. pava Boras and Paulina Boras never meet: the two strings share two letters and no token, so containment fails and the three-letter-prefix rule fails with it. They are one woman. Born 24 December 1895, the same parents, and both married an Ivan Krmpotić — one copy carrying the marriage, 1917 at Senj, and no children; the other seven children and no marriage date. That is precisely the split Danijel Kalanj shows, and it was invisible. So a pass was added that never looks at the name. It starts from the family: children of one couple who share an exact birth date. On its own that proves nothing — this tree holds 113 such pairs and most of them are twins, several labelled so in the tree itself. The discriminator is the marriage. Twins do not marry the same person; a record entered twice does. And it has to be the spouse's NAME, not their record. Matching on ids found one pair and missed the Borases, because the tree holds two Ivan Krmpotićs — duplicating a wife duplicates her husband along with her. On names it fires eight times across 15,643 people, with no twin among them. Seven are married-in South African and Welsh lines; the Borases are the only one in the Croatian material, which is a statement about where the tree was built carelessly, not about where this pass works. Two tiers were being counted and never shown. `same day and same parents` — the five groups this row added on the first sitting — was never added to `/duplicates`, so the page's total said thirteen while its tables listed eight. Both new tiers are on the page now. A tier the reader cannot see is worse than no tier at all. --- A third sitting, 17 September 2026, and the blind spot this time was chronological. Every test the tool had was built on a birth: the birth day, the birth year, the birth place. Nothing anywhere in it looked at the death. That had already been written down as a problem and left — this row's own note says of Danijel Kalanj, *«the plainest of all without qualifying»*, that both his records carry the same death date, 26 June 1876, and he sat in the bare «same day» tier anyway because nothing could see it. Two changes, and they are the same idea applied twice. A confirmation, `same day, and the same death day`. One person dies once and on one day. It is a confirmation of the same weight as *same parents* — and better in one way, because siblings share parents and cannot share a death day. It fires once, on Danijel Kalanj, which is exactly the case the row had already identified and could not grade. And the name-blind pass now runs on the death date too. Keying on the birth day finds a record split down the middle when both halves kept the birth. It cannot find Louis Barry, who is in this tree twice with two different birth years — 17 February 1920 and 1907 — so no key built from a name and a year could ever put his two records in one bucket. What they share is the day he died: 20 March 1983 at Cape Town, both born at Elliot, both married to Louise Sophie van Heerden. Live groups 31 → 32. One veto, and it earned its place immediately. Two records that both carry a full birth date and disagree about it are two people whatever else they share. Elizabeth Margaretha Catharina Botha (b. 29 January 1797) and Margaretha Debora Maria Botha (b. 15 May 1802) are sisters who both married a George Aldric and are both entered as dying on 15 March 1889. Without the veto the pass would have announced her as one woman. She is two, and the coincidence is recorded rather than resolved. And a latent false claim closed on the way past. `full_date` tested a date by its *length*, so «FEB 1967» passed as a day and would have been graded «*same day. Near-conclusive.*» No group published here ever rested on one — checked before the rule was tightened — so nothing was re-graded. The door is shut. Both tiers are on /duplicates. That rule was learned last sitting and is not being broken this one. --- 20 September 2026 — the index reaches the duplicates at last, and it answers two of them. A live group is two tree records that may be one person, and the register settles it with one question: is there one baptism entry, or two? Four of the seventeen fall inside the Otočac index's window of 1834–1858. Andreas Žubrinić, 1844 — one entry, and one of the two records is a fragment. The tree has `andreas-zubrinic` (13 Nov 1844, Otočac, Michael × Catharina) and `andreas-zubrinic-xubrinich` (Nov 1844, no day, no place, same two parents' given names). The index holds exactly one baptism — 13 November 1844, «Michaël Žubrinić × Catharina», `1:1:QKMK-S6M9`. One copy has the day and the place and the other has neither: the shape of Danijel Kalanj and of pava / Paulina Boras, one record entered twice. Alexander Žubrinić, 1839 — and the register does not give the mother a surname. This row has said since the 15th that these are *«two boys, born the same day, to two different Michaels»*, and `duplicates.py` grew `parent_surnames()` precisely so that *Orešković* and *Draženović* would keep them apart. Those two surnames are not in the register. The index holds one baptism — 14 March 1839, «Michael Žubrinić × Magdalena», `1:1:QKMK-STYY` — and the mother has no surname in it. Whatever supplied *Orešković* to one record and *Draženović* to the other, it was not this baptism. So the discriminator is working on data that is not evidence. That does not make them one boy. It means the reason given for their being two does not come from a document, and this row should stop repeating it as though it did. And the caution is the same one that keeps recurring: the index is not complete. The measure is Karlobag, where the entire Luckinić holding is seven entries for a family documented across generations, and ahnentafel 105's own baptism is missing while her father's other children are there. «The index holds one entry» is not «the register holds one entry» — though in the Alexander case it does something stronger than point: it removes the grounds for the opposite conclusion. [notes/the-index-and-the-duplicates.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-index-and-the-duplicates.md) Two of the four were not answered: *Lucia 1835* returns nothing at Otočac, and *Magdalena 1847* returns one entry to Philippus × Maria Begović, which is neither of the pair. --- And then the index found four groups that are one household — which exposed a tier this tool never had. Krivi Put holds one couple, Antonius Blažević × Maria Šojat, producing exactly the four names that make four separate duplicate groups: Ivan 10 Aug 1853, Mile 8 Sep 1855, Marko 27 Mar 1863, Ane 30 Oct 1865 — one indexed baptism each. In the tree those four are nine records, and every one names the same two parents under three spellings apiece — *Antona / Antone / Anton Kljapo Blažević* and *Marija Sojat / Mare Sojat / Marija Šojat*. `parents_agree()` already returns True for all four. They could not reach the confirming tier for one reason only: it requires a full birth date on EVERY record, and in each group one copy carries a year and nothing else. That is the commonest shape a split record takes — one complete copy with a day, a village and sometimes a house number, and one stub with a year and two parents. Danijel Kalanj is that shape. So is pava Boras. So is Andreas Žubrinić. And it was invisible, scattered across the two weakest tiers where the fact that the parents agree could not be seen. New tier: «same year, same parents, and one record is a fragment». It catches nine groups — *same year and place* falls 8 → 2 and *same year only* 8 → 5. Four Blažević of Mrzli Dol 6, two Kalanj of Klenovica (Lovre/Lovro × Ana Pavelić), three Žubrinić (Philippus × Maria Begović). Not one of the nine is a plausible sibling pair: in every case the stub carries no date detail of its own at all. Six of the nine have index support. Ivan, Mile, Marko and Ane as above; Andreas 1844 and Magdalena 9 May 1847 each return exactly one baptism, and the 1847 entry's parents — *Philippus Žubrinich × Maria Begovich* — match one of the two tree records to the letter. It is ranked below the same-day tiers and it is not a confirmation. Two siblings can share a birth year and two parents, and a year without a day cannot rule that out — which is why the tier is named for what it observes rather than for what it concludes. It is on /duplicates, because a tier the reader cannot see is worse than no tier at all. And a veto the tool was missing, which is the same idea the other way round. It had *«rejected — two death years»* — a necronym is two children, because one person dies once. It had nothing for births. Magdalena Kalanj was being carried as a live duplicate: 8 June 1829 at Klenovica in one record, 19 February 1829 at Ledenice in the other. A different day, a different month and a different village — and the only reason she sat in *«same year only»* rather than in the rejections is that the tool never looked at whether two birth days could both be true. «rejected — two birth days» now does, and it is the stronger half: not two death *years* but two birth *days*, both written out in full and both unhedged. It fires once. A group reaches it only when every record carries a full date, so a complete copy paired with a year-only stub is untouched — that is the fragment tier's business, not this one. Live groups 32 → 31; rejections 28 → 29. Both vetoes are named on /duplicates with what they mean, because *«rejected» means the evidence for a duplicate failed, not that two people are proved distinct* — and these two are the exception: they are positive evidence of two people. --- 21 September 2026 — the seven weakest groups were taken by hand, and two of them turned into vetoes the tool can apply itself. Of 31 live groups, twenty-four sit in tiers that already confirm them. The residue was seven, in the tiers that say least: *same year only*, *same year and place*, *same day*. Every one was read out in full — parents, spouses, children, places — and two were decidable. `rejected — different parents by surname`. `parent_surnames()` has existed since 15 September and was used only to withhold confirmation. That left the very case it was written for sitting in limbo: the two Alexander Žubrinić of 14 March 1839 have a Michael for a father in both and mothers reducing to *magdalena* — one Orešković, one Draženović — so `parents_agree()` correctly refused to call them one boy, and nothing then called them two. The group stayed «live», which is the tool saying *I do not know* about something this row had already decided in writing. A disagreement good enough to withhold a confirmation is good enough to reject. `rejected — one died a child and the other married`. Thomas Žubrinić, born 1822: one record dies in 1829, aged seven, with no spouse and no children; the other has a wife, Joanna, and a son, Nicolaus. Those are two people, and no amount of name-matching says so — the parents veto needs parents and neither record here has both. What separates them is a life that ended before the other one's began. Each new veto fires exactly once across the whole tree, which is the property that matters: a veto that over-fires hides a real duplicate. The confirming tiers are untouched at 24. Live 31 → 29. ## The five that stay live, and why that is the right answer Three are one person split down the middle — the shape this row already named as «the plainest of all» for Danijel Kalanj, where one copy carries the parents and the other the marriage. Magdalena Žubrinić 1812 (parents *Antonius × Maria* in one, husband *Wolfgangus Adamić* and three children in the other), Marija Kalanj 1857 (parents *Ivan Kalanj × Josipa Miletić* in one, husband *Tomo Miletić* in the other), Michael Žubrinić 1802 (father *Jure* and a wife *Helena* in one, a wife *Margarita* and three children in the other). Nothing contradicts, and nothing confirms: Danijel was settled by a shared death date and none of these three has one. They stay live honestly. Two turn on a second marriage. Joso Blažević 1843 has *the same parents in both copies* and two different wives, Kate Šojat and Ruže Prpić; Josip Prpić 1880 has a mother in one copy and both parents in the other, compatible. A man who remarried and a pair of brothers who shared a name are both ordinary in these villages, and this archive has met the second often enough — the seven Michaels, the two Georgius Urode — not to guess. A register settles it and a tier cannot. --- 28 September 2026 — triaged, and one block of five is settled. The file now reports 29 live groups and 33 rejected of 62. The 29 are not one problem: * Nine are South African — Barry, Boshoff, Davies, Fick, Kolbe, Pretorius, Reynders, van der Merwe — in the top tier *«same day, same parents, same spouse»*. Hedviga died in Johannesburg and the tree carries that side; these are almost certainly real duplicates and none of them is a person this archive researches. They are somebody's to fix and not this archive's to guess at. * Nine are «same year, same parents, and one record is a fragment», which is the tier that needs a document rather than a rule. * Five are Žubrinić and Zubrinich, and those are settled below. THE FIVE, AND THE EVIDENCE IS THE SAME EVERY TIME. | Who | Born | Both records | |---|---|---| | Catharina Žubrinić | 20 Nov 1840, Otočac | no source on either · 2 events each | | Josephus Žubrinić | 16 Mar 1845, Otočac | no source on either · 2 events each | | Francisca Žubrinić | 6 Mar 1846, Otočac | no source on either · 2 events each | | Nicolaus Žubrinić | 3 Aug 1847 | one entered «Nicolaus Žubrinić», the other «Nicolaus Žubrinić Xubrinich» | | Katalyn / Kate Zubrinich | 26 Jul 1886, Port Pirie | one source each · *Katalyn* and *Kate* are the same name | Identical full birth dates, identical parents, no source on either side, and in two cases a name variant rather than a second name. Four of the five are consecutive children of one Otočac household in the 1840s, each entered twice. This is a defect in the MyHeritage export, not in the family, and the method here is to report the tree rather than edit it — so they stay, doubled and labelled, and the count of published people carries them. What would close them properly is the Otočac baptismal register for 1840–1847, which would give one entry per child and end the argument. It is the same book row 27 wants. This is a triage and it says so. Nine South African groups have been looked at and left to the family that owns them; nine fragment-tier groups are untouched. Only the five above are adjudicated.

The by-name file was built out of Selce and its neighbours, where the connectors are *aliter · dictus · dicti* and the scribe underlines the pair. Karlobag, found on 15 September, uses *vulgò* and does not underline. So the sweep that produced 17 phrases was looking for the wrong shape in every parish except one. Re-run it across the pages already read with the connector list widened — *vulgò · alias · rečeni · zvani* — and against the surnames the tree carries as double-barrelled, which is where a by-name hides once it has been absorbed (ahnentafel 106 «Smojver Vukić» was exactly that). --- Re-run 15 September 2026 across every transcription on disk — `readings.psv`, `found.psv`, `searched.json`, all of `notes/`, the film maps and the grave surveys — with the connector list widened to *aliter · alias · dictus · dicti · dicta · vulgò · rečeni · zvani · vide · ut ibi · cognomine · nuncupatus*. 123 connector phrases in 54 files. Almost all of them were already in the file, which is the answer the row wanted: the narrow sweep had not in fact missed a parish, because the Karlobag *vulgò* pairs were caught by hand the same day they were read. Three things came out of it. One new pair. «*Verbanovich Margarita … vide Urbanovich … 15 9bris*» — Verbanović ⟷ Urbanović, on the same KAZALO page as «*Vukich vide Smojver*». It had been recorded in the film map and in the register and never added to the by-name list, which is exactly the gap a sweep exists to close. One false friend, and it matters more than the pair. In the Otočac *status animarum*, «*dictus / dicti*» is not a connector at all — it is the Latin for *«of the aforesaid»*. «*Michael filius dictus Georgij*» is Michael son of the said George; «*Marcus et Josephus … Georgii dicti Gemelli*» is the twins of the said George, not a household called Gemelli. **A sweep that harvests every *dicti* invents by-names out of pronouns.** The test — *does the word sit between two names* — is now written at the top of `bynames.psv` so the next sweep does not have to rediscover it. And one lead that cannot be closed from disk. The Karlobag marriages of 1806, folio 3, list a groom as «*Vulgo Zalic*» — a *vulgò* with the surname it attaches to missing from the summary table. It needs that page reopened, not another sweep. Still not in the list, and correctly so: «*Georgius Saldatich aliter Venesarich*» and «*Catharina Polich aliter Antich*», both on image 304 of the Selce deaths. They were read at half scale and reading is reading.

Done in nineteen HTTP calls, 15 September 2026, and the answer is better than the audit this row planned. The row was going to open every film and guess. It did not need to: `/service/search/catalog/item/koha:<n>` returns one `film_note` entry per ITEM, each with its `digital_film_no`, its physical roll number and its item range. Nineteen calls, eighty-two films, written verbatim to `sources/familysearch/filmnotes.psv` and checked against a second fetch by SHA-256 before being committed. «A film number is not a book» is now explained, not just observed. The catalogue's unit is the *item*; the viewer's unit is the *roll*. A roll carries several items and the items on one roll need not come from one parish. > **Karlobag's baptisms 1691–1804 are DGS 5481648 — *items 4 to 5 of roll 1419107*. Items 1 to 3 of that same roll are the parish of KAPTOL, `koha:655210`, three hundred kilometres inland. So `/search/film/005481648` opens the roll at item 1 and lands in Kaptol. That is exactly what happened here, and it cost a sitting. Thirty-one of the eighty-two films do not begin at item 1, across seventeen of the nineteen parishes**. Trsat's oldest register is *item 8*. Bribir's is *item 7*. Those are the films a DGS number cannot reach, and they must be opened by waypoint. The same shape explains Otočac. DGS 5482656 is *item 6* of roll 1536689 — five items of something else come first — and 5482657 is catalogued as births *and* deaths because it is *items 1–2*, two books, which is why walking it gives 483 images of baptisms and then END OF ROLL. And a book can have two film numbers. Ledenice's registers were filmed twice: roll 1419155 (DGS 5482218) and roll 2119536 (DGS 5498069) are the same three books, decades apart. Neither is wrong. `holdings.psv` had only the first. One real gap closed. Brinje sat in `holdings.psv` as *«film notes not pulled»* with no DGS number at all — and nothing noticed, because nothing compared the file with its source. It is DGS 5498262 (Rođeni 1888-1900, Vjenčani 1888-1899) and 5498263 (Umrli 1888-1899). That is row 50 answered as a side effect. `tools/holdingscheck.py` now gates the build so the summary can never again claim a year the catalogue does not have.

Answered 15 September 2026, and it re-orders the rows above it. Eight surnames of this family were swept through the index and tallied by place and event type; the result is in `sources/familysearch/index-coverage.psv`. Indexed for baptisms only — these are walks, not lookups: Otočac (165 baptisms, not one death), Selce (77, confirmed twice), Karlobag (four surnames tested, every hit a baptism, range 1717–1861), Krmpote-Vodice. Deaths are indexed: Senj — also under the Latin place string *Senia*, which a search for «Senj» misses — Krivi Put, Novi Vinodolski, and thinly Crikvenica. This corrected row 23 before it was started. That row was written asserting the Karlobag death index reached ten ancestors; the earlier evidence for that was a Karlobag date range of 1717–1861 which turns out to be entirely baptisms. Row 23 is a walk through the book's own KAZALO, not an index lookup — and the KAZALO is ordered by folio, not alphabetically, so a surname is found by scanning, not by jumping to a letter. And a trap worth more than the map. Ahnentafel 27 is indexed under Pilipić, her married name — a search for *Uroda* returns nothing at Senj. A married woman's death is indexed under her husband's surname. Every sweep for a maiden name will walk past her.

`targets.json` carries 25 unplacedVillages — place strings from the tree that no parish in `parishes.psv` claims. Each one is either a spelling the gazetteer does not fold, a village in a parish not yet in the table (which is how Bribir and Sv. Juraj were added on 14 September), or a place outside this family's districts. Until they are placed, the people standing on them cannot be routed to a book. --- Triaged 15 September 2026, and the list is mostly not what it looked like. Eleven are emigration destinations — Port Pirie, Quorn, Port Germein, Whyalla, Yarraville, Marrickville, Sydney, Kurri Kurri, Arizona, Johannesburg, and «South Australia» itself. Nine are cities and regions outside these districts — Zagreb, Osijek, Pula, Belgrade, Garešnica, Fužine, Sveti Križ, Tomašica, «Yugoslavia». No parish table should claim any of them and none is a defect. Two were data artefacts and are fixed at the tool. * «"Croatia, Church Books, 1516-1994, " database with images…» was the *birthplace* of three people — a source citation pasted into a place field. A gazetteer that parses it invents a village. `parishlib` now recognises a citation and returns no place, which is the truth. * «Montani Stanište» with two spaces was a different village from «Montani Stanište» with one, so ahnentafel 111 and her own father stood in two places that are one place. `norm()` now collapses internal whitespace. And three are real Croatian places, behind four direct ancestors. | Place | Who stands on it | |---|---| | Smiljan, Općina Gospić | ahnentafel 426, Simoni Pekass | | Montani Stanište | ahnentafel 111, Lucia Tomljanović, and ahnentafel 222, her father Pauli | Those are not gaps in the record. They are gaps in the gazetteer, and the difference now shows: see row 45 and row 56, which carries the job of establishing which parish each belongs to.

Done, 15 September 2026, and the answer was not the one this row expected. The walls are drawn on our line. `walls.json` was byte-identical at every commit this week — 9 unplaced, 7 nameless, 6 outside, 4 reachable — while nine direct ancestors were read. That looked like a correct null and was not. `walls.py` could not see the reading. It loads `ancestors.json` and `placefixes.json` and nothing else: the parish table tells it what is filmed, and nothing told it what had been read. So it printed «*This wall is reachable and has not been opened*» on four walls, and for three of them the register had been opened by this archive that same morning. And one wall was false. Ahnentafel 220, Nicolaus Gerkacs, read «*No place is recorded for this household anywhere on the line, so there is no register to open. The wall here is a place, not a name.*» His death entry puts him in his son Antonius's house at Karlobag. The tree has no birthplace for him; a document does. Both fixed at the tool. `walls.py` now reads `sources/readings.psv` and says which documents are held and which book is still shut. `sources/walls.psv` is created with a format — `ahn|field|value` — where `place` supplies a place read off a document and `why` cites it; a place does *not* assert a grade, it is fed back through the normal grading so the kind stays derived. **Ahnentafel 220 moves *unplaced* → *reachable*. 8 unplaced now, 5 reachable. And the re-grade names the next job. Ahnentafel 104, 108, 212 and 220 all want the same book — Karlobag baptisms, 1691–1861, filmed, and nobody has opened it for any of them.** Four walls, one sitting, and it is the highest-yield unopened book on the chart.

**The first job is done, 15 September 2026: `sources/graves/*.psv` is wired in and published at /stones/. `tools/graves.py` now reads the photographed surveys beside the Find a Grave harvest and keeps the two kinds of evidence apart by construction — every top-level key of `graves.json` still counts the index** and nothing else, so Port Pirie's «*72 memorials and not one of them is in Croatia*» stays true. The surveys live under `stones`. 226 names, 137 of them `r`. Otočac 32, Senj 194. 16 match a person here and 66 stones carry more than one name, which is the same household claim a plot number makes. Two things the wiring taught that the files alone did not. * A Croatian stone gives a married woman under her husband's surname. Matching on the maiden name alone missed ahnentafel 3, 5 and 7 outright. The matcher now tries `rel.spouses` too, and marks those matches as made through a marriage. That is three direct ancestors recovered by one rule. * The ±1 birth-year drift is not a Senj quirk and it does not run one way. Putting the disagreements side by side surfaced a third — Franjo Žubrinić at Otočac, 1846 on the stone for 1845 in the tree — and showed that an earlier note's claim that the stones were *consistently* a year early was wrong. Ljubomir's is early; the other two are late. Corrected in the note, the register and the page. Three disagreements, all births, all one year, against not one death year in dispute in 226 names. The matcher reuses the archive's own vocabulary — `givennames.json` already knew that Stjepan is Stephannes and that Franjo is not Franciska, which is what stopped the first draft joining a man to a woman. Only an `r` row may contradict the tree: a `p` digit is not evidence and has been wrong four times. What is left is the original second job. 56 of the 72 Find a Grave memorials still match nobody, and 210 of the 226 stone names likewise. Some are collateral lines, some are the *found* people of row 47, some are strangers. The house numbers from row 1 give a handle at Otočac; at Senj the surnames are this family's — Prpić, Špalj, Papić, Vukelić, Tomljanović, Boras. --- The matcher was reading the register file, not the tree — and it was the fifth tool in this archive caught doing it. `graves.py` matched stones against `people.json` and nothing else. That is not the tree, and it is not even everyone with a page: fifty-two direct ancestors sit in `ancestors.json` and not in it, so their graves could not match by construction. Beyond both is everyone who exists only as a name on somebody else's record — a parent, a spouse, a child — with no page and no slug. Sixteen matches became twenty-one, and every new one is a person with no page. Martin, Lucija and Kata Špalj; Nikola Šojat; and Ivan Defrančeski, 1925–1995, Hedviga's husband — who now matches his own gravestone because the matcher was taught where to look, not because somebody wrote a correction by hand. That is the difference between fixing an instance and fixing a tool, and this archive had been doing the first. One person turned out to be claimed by two stones. Two Senj stones read KATA ŠPALJ — *1884–1975* and *1886–1981* — and the tree holds exactly one Kate Špalj, born 1886, with no death recorded. Both matched her, because a match needs a year within two and there was no death year to disagree with. The better-scoring stone keeps the match, the other is unmatched again, a tie drops both, and the contest is published either way. One of those two women is not in this tree, and publishing her under somebody else's name would be the Defrančeski mistake run backwards: not a person missed, a person invented. What is left is what was always left: 56 Find a Grave memorials and 205 stone names matching nobody. The pool they are matched against is now the right pool, which is the precondition for the rest. --- 20 September 2026 — and it happened a sixth time, to the file written to prevent it. `tools/graves.py` was taught on 15 September to match against the tree rather than the register file, and the note above calls that *the fifth tool in this archive caught doing it*. It was still not the right pool. `sources/found.psv` was not in it — and that file exists for one reason: a person who is in a register and not in the tree had nowhere to be recorded. It holds the single population a stone is most likely to name and the tree least likely to explain. The cost was one name, and it is the name that proves the point. «MATILDA PAPIĆ 1895–1921» stands third on a Senj cross. She was photographed, read at full resolution, found to be absent from the tree, and written down in `found.psv` as a discovery — the archive's own record of its own finding. Then the matcher ran and reported her stone as matching nobody, because *the file written to hold her was not the file she was matched against*. Both her years agree to the digit. Matched stone names 21 → 22, and nothing else moved — the other six found people who share a surname with an unmatched stone are Žubrinići at Otočac, and not one of them agrees on a given name or a year. The pass is conservative and stayed conservative. She keeps her null slug, and the page now says what she is. The row would otherwise have printed «*in the tree, with no page of their own*», which is the one thing she is not. A found person stands beside the tree — that separation is what `found.psv` is for, and a matcher that quietly dissolved it would have been worse than one that could not see her. What is still left is what was always left, and it is not a matching problem: 204 stone names and 56 memorials match nobody because this archive does not hold those people. The surnames are the family's — 45 Prpić, 21 Papić, 20 Žubrinić, 19 Špalj, 18 Vukelić — so they are collateral lines and strangers, and only a register separates the two. --- CLOSED 21 September 2026. This row's own task was the matcher — «the stones are matched against the right pool at last» — and the matcher is finished: the surveys are wired in and published, `graves.py` keeps the photographed stones and the Find a Grave index apart by construction, the found people are in the pool, and Matilda Papić is matched with the page saying what she is. What remains is not a matching problem and never was. 204 stone names and 56 memorials match nobody because this archive does not hold those people — 45 Prpić, 21 Papić, 20 Žubrinić, 19 Špalj, 18 Vukelić, all family surnames, all collateral lines and strangers. Only a register separates the two, and that work lives in the reading rows, not here. Leaving this open would misreport it. A row that stays «running» because something adjacent is unfinished is the same fault as row 48's electoral rolls, which reported eleven of nineteen read for days after all nineteen were done.

Done, 15 September 2026, and it caught the thing it was written to catch. The row said each of the found people needed re-testing against parent and spouse references, not just published names, because this archive had already been wrong that way once. It was wrong that way again, and worse. Four rows came out of `found.psv`. * Thomas Žubrinić, 1846 — published, and both parents match: Michael Žubrinić × Helena Orešković. * Luce Žubrinić, 1850 — published, both parents match: Mio × Jele Orešković. * Helena Orešković — in the tree as a parent reference with no page of her own, which is precisely the Karlobag-wives case repeating. * IVAN DEFRANČESKI, 1925–1995 — recorded from the Senj gravestone as *«the surname does not occur once in this archive's tree»*. He is Hedviga's husband. The tree carries *Ivan Anton Defrančeski, b. 1925 d. 1995*, on her own marriage, with a null slug because he is not published — so a search of `people.json` returns nothing. Both dates match the stone exactly. The archive photographed the grave of the man at its centre and filed him as a stranger for a day. The lesson was already written down and did not stop it happening, so it is a gate now. `tools/foundcheck.py` tests every `found.psv` row against every name in the tree — published people and every `rel.parents`, `rel.spouses`, `rel.siblings` and `rel.children` reference — two ways: *name + year within one*, and *name + BOTH parents agreeing*. A match a person has looked at and dismissed goes in a `SETTLED` table with the reason; anything else fails the build. It runs as `check:found`. Eighteen people remain in `found.psv` — the number this row started with, arrived at honestly. Three still resemble somebody in the tree and all three are settled in writing: *Maria Rosaria Antić* (ahnentafel 24 has no children published at all), *Maria Magdalena Theresia* and *Maria Žubrinić* (the parents fail the test), and Matilda Papić, who probably IS Franka or Sulka — the open question row 53 left standing, not an error.

Both are in the register as *pending* and neither can move without a decision that is not this archive's to make. FindMyPast — Port Pirie electoral rolls: eleven of nineteen read, and the trial expired before the rest; the rolls give an address and an occupation year by year, which is how a household is watched over time. Southern Rhodesian immigration 1953–65: an entire country known from two words on a later record, and no route into it has been identified at all — which is a different thing from having searched and found nothing. Both want either a subscription or a decision to let them stand as honest nulls. --- 21 September 2026 — the electoral-roll half of this row is CLOSED, and has been for days without the row saying so. It reads «eleven of nineteen read, and the trial expired before the rest». The month's subscription finished them: `sources/findmypast/electoral-rolls-2026-09-16.psv` holds fifty transcripts for nineteen people across five roll years — 1913 and 1935 in New South Wales, 1939, 1941 and 1943 in South Australia — each with an address, an occupation and the household at that address. They are published, with the rest of the catch, at the Catch Abroad. What this row wanted — a household watched over time — is exactly what an address repeated across roll years gives, and the page marks how many do. The Southern Rhodesian half is untouched and still has no route in, which is why this row stays open rather than closing. Splitting the two would have made that visible sooner; a row with two threads reports the state of whichever one somebody last wrote about.

Added 15 September 2026 because the set was not visible. Four of the five letters had rows here — 10, 11, 12 and 13 — and the fifth had none, which is the one that generated the other three. A reader of this list would have counted four letters. Državni arhiv u Rijeci, Ispostava u Senju — sent 13 September, answered 14 September by Denis Rukavina: «*Sabirni centar u Senju nema u svom posjedu arhivsku građu kojom bi se moglo udovoljiti Vašem zahtjevu*» — the Senj collection centre does not hold it — and he named three doors in its place: Rijeka, Gospić, and the competent parish offices. That reply is why rows 10, 11 and 12 exist. It was a nil return that did more work than most positive ones. | To | Sent | Where it stands | |---|---|---| | DAR, Ispostava u Senju | 13 Sep | Answered 14 Sep — nil return, named three further doors | | Državni arhiv u Gospiću | 13 Sep | Acknowledged 14 Sep, queued — row 13 | | Državni arhiv u Rijeci | 15 Sep | Answered 15 Sep, refused — row 10 | | Župni ured Senj | 15 Sep | no reply — row 11 | | Župa Krmpote, via Ledenice | 15 Sep | Answered 15 Sep — the registers left the parish — row 12 | Three of the three doors Senj named have now been tried, and four of the five letters have answers. Only the Župni ured Senj is still out. And the reasoning that produced these letters has been half disproved. It was written here that *«if anything comes, it is most likely to come from a parish priest, because the books the films do not cover are the ones that never left parish custody»*. Krmpote answered the same day: its registers went to the Rijeka Archdiocese, and from there to the State Archive. The parish has no contact with its own older books. That is worth leaving on the record as written and then corrected, because it is the assumption that shaped where five letters went. Full text of every reply is in `requests/REPLIES.md`; the letters as sent are in `requests/sent/`.

Found by the missing-row gate on 15 September 2026. It had been *pending* in the register since the 14th and had no row here at all. Not one page of the Brinje register has ever been opened, and Brinje is not in this archive's gazetteer. It got onto the list the hard way: a 1910 American manifest gave a Kalanj's residence as *Brinje*, which sent someone back to re-read a Latin baptism of 1864 — *«Brinjensis parochiae ruricolae»*, country people of the parish of Brinje, forty kilometres inland from the coast where this archive had always put them. Three surnames meet there — Kalanj, Perković and Borić. The films are catalogue-verified for 1888–1900 across births, marriages and deaths, which is late for an 1864 baptism but right for the generation that emigrated. The Gospić letter also asks for the Kalanj × Perković marriage at Brinje, so row 13 and this row may answer each other. The films are located, 15 September 2026. Row 42's catalogue sweep pulled the film notes Brinje was missing: DGS 5498262 — *Rođeni 1888-1900, Vjenčani 1888-1899*, items 1–4 of roll 2121763 — and DGS 5498263 — *Umrli 1888-1899*, item 1 of roll 2121764. Both begin at item 1, so both are reachable by film number, which is not true of thirty-one of this archive's eighty-two films. The dates are the problem and they always were. 1888 is twenty-four years after the baptism that sent us here, *«Brinjensis parochiae ruricolae»* of 1864. What these two films can carry is the Kalanj × Perković marriage the Gospić letter asks for, if it falls 1888-1899, and the generation that emigrated. Brinje is still not in the gazetteer and still has not been opened. --- 20 September 2026 — which books are indexed is now measured rather than assumed. `sources/familysearch/indexed.psv` holds the table, probed parish by parish through the persona search API. And the first version of that table was wrong, corrected the same day: marriages and deaths ARE indexed, through *Hungary, Croatia-Slavonia, Marriages 1678–1927* and *…Deaths 1681–1990*. They were invisible because the search is ranked and capped — a broad query returns about sixty rows and whole collections sit below the cut. Narrow by date and they appear. Brinje has nothing indexed, under Kalanj or Marić, for any event. Combined with row 57's finding that no Kalanj from Klenovica or Brinje filed a US naturalisation petition, this village is reachable only by opening its film — which is what this row said, and it is now measured from both directions. --- 29 September 2026 — opened, at David's own request, and logged in `sources/searched.json` as `fs/brinje-vjencani`. DGS 005498262 (waypoint 9R26-82S) is one film holding both Brinje registers; its Vjenčani (marriages) volume was read from its first page, images 493 to 511 of 693, entries roughly No. 1 through the mid-30s, 1888 into 1889. Two images (509, 512) loaded too small to read with confidence and were left rather than guessed. One Perković, so far, and he is a witness, not a party. Entry No. 7, 12 April 1888 — Grgo Božičić, a widower of Brinje-Križpolje, marries with witnesses Krznarić Luka and Perković Mate. The first confirmation of the surname from the register itself, independent of the emigration records that sent this archive here. No Kalanj and no Borić in this stretch, as party, parent, or witness. What is still closed. 664 of this film's 693 images remain: the rest of the Vjenčani book (images 512 to roughly 620, into 1899) and, before that, the two Rođeni volumes (images 1–492, 1888–1900) — where a Kalanj or Perković baptism, or a Perković godparent, is actually the more likely find, since Josephus Kalanj and Rozalija Perković's own 1864 marriage almost certainly predates this book entirely. DGS 005498263 (Umrli/deaths, 1888–1899) has not been opened at all. --- 29 September 2026, continued the same day — a Kalanj, at image 530, before the book even reaches 1891. Asked directly to keep walking. Entry no. 13: KALANJ NIKO, bachelor, 40, of Brinje house 231, son of KALANJ JURO and Ana, marries a widow recorded as Marija. Niko and his father Juro are the first Kalanj men this archive has found actually resident at Brinje — not the coast family remembered as being «of» the parish on the strength of one Latin phrase, but a household in the village itself, a generation before anyone here could otherwise reach. Neither is in the tree; both are now in `found.psv` and at `/found/niko-kalanj/`. The facing entry, same image, is three Borić men. Recorded as «Vid» Borić (the given name resists a confident reading), of Kamenica — a hamlet of Brinje parish — son of Borić Luka, with a second witness Borić Jozo. This page has said for a fortnight that a Jure Borić of Brinje gave a daughter to a Papić of Senj, sourced from a 1910 emigration manifest and never from a Croatian register. This is the first confirmation from the parish's own book that Borić is real and resident, at Kamenica specifically. Whether any of the three is kin to the already-named Jure Borić is not established. Recorded at `/found/vid-boric/`. Read so far: images 493–532 of 693, entries roughly No. 1 to the low 20s across two calendar years (1888–1890) — the book renumbers each entry from 1 at the start of every year, which cost one confused re-read before the pattern was seen. Image 533 loaded too broken to read (large black load-failure blocks) and wants a retry. --- 29 September 2026, continued further the same day — four more Borić men, images 533–539. Image 533 (retried) held entry no. 19, 7 July 1890: Borić Juko, widower, 38, of Brinje house 126 itself — not the Kamenica hamlet — son of Borić Ivan. Image 535: Borić Miko, widower, 55, of Križ Kamenica, born about 1835. Image 538, entry no. 28: Borić Niko, bachelor, 25, of Brinje house 142. That makes five Borić men across one 47-image stretch — Luka, Vid, Ivan, Juko, Miko, Niko — a real presence, not a single paper's claim. All recorded in `found.psv` with their own `/found/` pages; none in the tree. --- 29 September 2026, images 540–566 — a puzzle, and a second Kalanj household. Image 564: Borić Marija, spinster, 19, of Brinje house 142 — Niko Borić's own house — daughter of a Niko Borić, marries Hofjenc Niko. The father's name and house match the bachelor who married in 1890 exactly, and the ages do not comfortably fit one man; recorded as its own open puzzle, not smoothed over. Image 566: Kalanj Mijo, bachelor, 55 (born about 1838, the oldest-dated Kalanj this archive holds), of Brinje house 249 — a second Kalanj address, distinct from Niko and Juro's house 231 — witnessed by Kalanj Mate. Three named Kalanj men and four Borić men now, across four Brinje houses, independent of the coastal line and the single emigration manifest this page used to rest on. Images 493–566 of 693 read. One image (556) would not advance past its predecessor on three attempts and is logged unread. 127 images remain in this film, plus both Rođeni volumes (1–492) and the whole second film (Umrli, deaths). --- 29 September 2026, images 567–590 — a Perković as a party at last. Image 581: Perković Đuso, widow, 23, daughter of Perković Jozo, marries Vuković Marko — the first Perković found as bride or groom rather than a witness, and the exact surname the Gospić letter asked this archive to find at Brinje. Image 586: Borić Ive, son of Borić Jozo, of Križ Kamenica, marries Sertić Ane, witnessed by Borić Tomo — a sixth and seventh Borić man, a fourth Brinje-area location for the surname. One false-positive resemblance (Ive Borić vs. a coastal Pavelić) caught and settled. Images 493–590 of 693 read. 103 images remain in this marriage book, then both Rođeni volumes and the whole second film. --- 29 September 2026, images 591–622 — a likely sibling pair, found by house number. Image 595: Borić Luce, widow, 27, of Brinje house 126, daughter of Borić Ive, marries Sertić Tove. House 126 is exactly where Juko Borić (image 533) was recorded in 1890, son of «Borić Ivan» — the same name in short form. Recorded as a strong likelihood, not asserted as fact: no document names the two together. A caution earned in the same stretch: this «Borić Ive» cannot be the same man as the «Ive Borić» who married for the first time at Kamenica in July 1893 (image 586) — a father to a 27-year-old cannot be six months into his own first marriage. Two men, one name, kept separate. Images 493–622 of 693 read. The book turned over into 1895 in this stretch. One image (620) too blurred past the groom's bare name, logged for retry. 71 images remain in this marriage book, then both Rođeni volumes and the whole second film. --- 29 September 2026 — THE VJENČANI VOLUME IS FINISHED. Images 623–693 read; image 690 is the book's own «KRAJ» card from the Hrvatski državni arhiv u Zagrebu, the genuine last page. Three more finds in the final stretch: Miko Perković (image 653), a third Perković generation traced entirely from this register (Pere → Marko → Miko, at Novi Kut); a second, distinct Marija Borić (image 656, Kamenica house 24, kept apart from the first on purpose — the ages and houses don't fit one woman); Mile Borić (image 661, Križ Kamenica). The whole book: 3 Kalanj, 9 Borić, 3 Perković — 13 named people, none in the tree, all with their own `/found/` page. Ten images (509, 556, 620, 664, 671, 673, 675, 683, 685, 687) were too blurred or duplicated to read past a bare name and are logged unread rather than guessed. Next: the two Rođeni (births) volumes, images 1–492 of this same film — likelier than a marriage book opened in 1888 to reach Josephus Kalanj and Rozalija Perković's own generation. Then the whole second film (Umrli, deaths) remains untouched. --- 29 September 2026 — births opened, images 1–22 of 492, already four Kalanj/Borić finds. Juraj Borić (image 12), son of Tome and Jela Borić — a Borić man married to a Borić woman, a fifth Brinje house for the surname. Two women named Ane Kalanj, kept deliberately separate: one married to Grgo Dražičević (image 13), one to Viol Krmarić (image 18), both giving birth within the same three months of 1888. The first Kalanj *women* this archive has found at Brinje — everything in the marriage book was men — and roughly the right generation to be kin to Josephus Kalanj, whose daughter's 1864 baptism started this whole line of work. 470 images of this volume remain, then a second Rođeni volume, then the second film entire. Logged separately in `searched.json` as `fs/brinje-rodjeni`. Continuing. --- 29 September 2026, continued — images 23 to 81, no new finds. A long clean stretch: 59 images, several near-miss surnames (Kalafatić, Kalauz, Perić, Perko, and a mother's name misread in passing as Perković that a closer reread showed to be Perhorić) checked and ruled out rather than waved past. 411 images of this volume remain. --- 29 September 2026, images 82 to 112 — three new Borić finds. Stjepan Borić (image 88), born 4 February 1890 at Brinje house 28, son of a father of the same name. Marija Borić (image 89), a third Borić woman married out of the surname, named only as the maiden surname on her son Tomo Vičić's birth. Ivo Borić (image 106), of Brinje house 191, son of Ivan Borić and a mother *also* surnamed Borić — a second Borić-marries-Borić household, distinct from the one at house 141. Two more near-misses caught on closer zoom: a witness surname read at first as Borić turned out to be Dorić, and Perkić (met twice now) stays distinct from Perković. 380 images of this volume remain. --- 29 September 2026, images 113 to 122 — a Borić sibling pair and a third Kalanj house. Petar Borić (image 113), born 15 October 1890 at Brinje house 141 to the same couple, Tome and Jela Borić, already recorded as parents of Juraj Borić (1888) — confirming a sibling pair. Viol Kalanj (image 116), of Brinje house 118, married to Zoričić Kate — a third Kalanj house at Brinje, distinct from the two already found in the marriage book. A complication flagged rather than resolved: the same image also records Viol Krmarić, whose wife was named Kalanj Ane at image 18, here married to a woman recorded as Samson Ove — not yet explained. 371 images of this volume remain. --- 29 September 2026, images 123 to 134 — no new finds. 12 images, late April into August 1891. Three near-misses recur and are confirmed distinct once more: Kalauz, Perhorić, and Perkić. 358 images of this volume remain. --- 29 September 2026, images 135 to 140 — a likely father found for a puzzle already on this page. Marija Borić (image 137), a fourth Borić woman married out of the surname. Viol Borić (image 140), born September 1891 at Brinje house 142 to Niko Borić and Marica Bošnjić — almost certainly this Niko's first child, born the year after his 1890 marriage at the same house. That sharpens rather than resolves the puzzle already logged here: a Marija Borić who married out of house 142 in 1892-93 calling this Niko her father, an age gap this new birth makes harder still to credit. 352 images of this volume remain. --- 29 September 2026, images 141 to 147 — two more Borić women married out, one at a familiar address. Ove Borić (image 141) and Manole Borić (image 143) both married men of other surnames; Manole's entry gives Kamenica house 24, the same house this page already holds for a second Niko Borić's daughter who married out in 1897. Flagged, not asserted, as a possible link between two Kamenica Borić entries otherwise treated as separate. Also asked directly, 30 September 2026, whether the roughly thirty Kalanj/Perković/Borić people found at Brinje connect to each other: mostly not, beyond the sibling pairs and this father-son link already logged — see the Kalanj page for the full accounting. 345 images of this volume remain. --- 29 September 2026, images 148 to 156 — one find. Luce Borić (image 152), an unmarried mother at Kamenica, her son entered under her own surname - kept separate from the other Luce Borić already in this file. 336 images of this volume remain. --- 30 September 2026, images 157 to 159 — no new finds. Two entry-number gaps in this stretch (159, 167) were checked directly and read as genuine skips in the register's own numbering, not missed content. 333 images of this volume remain. --- Correction, same day: entry no. 167 was not a skip - it was on the page the whole time, missed on an earlier narrow pass; no. 159 has not been re-checked and should be treated as unconfirmed rather than repeated as fact. 30 September 2026, images 160 to 165 — a likely third sibling in the house-141 Borić family. Matija Borić (image 165), born December 1892 to the same Tome Borić already logged as father of Juraj (1888) and Petar (1890) at that house - probably a third sibling, though the mother's name here reads differently than recorded before. Also flagged: a second "Vičić Tomo" birth (image 160) to the same-named parents as one already in this file, twenty-two months apart - kept as written, not merged. 327 images of this volume remain. --- 30 September 2026, images 166 to 173 — no new finds. One apparent Kalanj sighting resolved by cross-checking parents' names already on record: the same recurring Kalauj family (image 158), not Kalanj. 319 images of this volume remain. --- 30 September 2026, images 174 to 179 — a Borić name at a new location. Ove Borić (image 179), married out to a Lučane household - the first Borić this file has placed at Lučane rather than Brinje, Kamenica, or Križ Kamenica. 313 images of this volume remain. --- 30 September 2026 — THE FIRST RODJENI VOLUME IS FINISHED, and it was shorter than this row assumed. Images 180-193 held no new finds; image 194 is the film's own «END OF ITEM» card, and this Rođeni item turns out to be catalogued 1888-1892, not 1888-1900 - a separate, shorter item from the Vjenčani book. Item 2, covering 1893-1897, begins immediately after on the same film and waypoint. This item's own tally: 4 Kalanj, 12 Borić, 0 Perković, plus the sibling pairs and intermarriages already logged. Continuing straight into item 2. --- 30 September 2026, item 2 opened — a likely sibling at Kamenica house 24. Kate Borić (image 201), born February 1893 to Niko Borić and Radotić Dace at the same house this page already holds for Marija Borić (married out 1897) - probably siblings, though the mother's name differs from Marija's record (Bare). Not merged; not resolved. --- 30 September 2026, images 202 to 206 — a possible new head of household at house 126. Stjepan Borić (image 203), son of Grgo Borić, born at the same Brinje house already logged for Ivan/Ive Borić's children Juko and Luce - but fathered by Grgo, not Ivan. Flagged, not merged. --- 30 September 2026, images 207 to 215 — two more Borić mentions at Kamenica. Tauka Borić (image 208), an eighth Borić woman married out. Jozo Borić (image 215), a witness - plausibly the same Jozo Borić already logged as father of Ive Borić (married 1893, also Kamenica), but not confirmed as one man by either document. --- 30 September 2026, images 216 to 218 — a confirmed second child for a Kalanj household already on this page. Franjo Kalanj (image 218), born August 1893 to Viol Kalanj and Kate Zoričić, the couple already logged here since 1890 - confirming a real household, not a one-off entry. The house number here (189) doesn't match the 118 recorded before; not resolved. Two more Borić witnesses at neighbouring house numbers, neither tied to a household. --- 30 September 2026, images 219 to 222 — no new finds. Two near-misses caught on closer zoom (a Zoričić misread as Borić, a Perkić misread as Perković). --- 30 September 2026, images 223 to 227 — the first proof a marriage-book puzzle became a real family. Hofjenc Niko (image 227), born October 1893 to Niko Hofjenc and Marija Borić - the exact couple already logged here from the marriage book (Marija Borić, daughter of Niko Borić house 142, married Hofjenc Niko c.1892-93). A child about a year later confirms the marriage actually took place. See the Kalanj page for the full write-up. --- 30 September 2026, images 228 to 229 — a caution earned, and a third birth at Kamenica house 24. A second, unrelated Niko Hofjenc turned up married to a Marija Hofjenc (not Borić) - caught before it could be conflated with the couple just confirmed. Then Luce Borić, daughter of Tome Borić and Marija Radotić, born at the same Kamenica house 24 already logged for two other Borić births this year - but under a different father's name and a different mother. Whether this is one large household or several sharing a number stays open. A Sertić family also holds that same house number, supporting the shared-address reading. --- 30 September 2026, images 230 to 235 — the cleanest confirmation yet. Anica Kalanj (image 235), daughter of Niko Kalanj and Marija Vučetić, at the very same Brinje house (231) and the very same Niko Kalanj whose 1890 marriage was this archive's first Brinje Kalanj find. No contradiction in the mother's name this time - a household confirmed cleanly. See the Kalanj page. --- 30 September 2026, images 236 to 239. A third surname (Hormanić) turned up at the Kamenica house-24 address already puzzling this archive, supporting a shared-address reading over one confused household. Two more Borić witnesses: Kate Borić (house 85) and Grgo Borić (house 74) — the latter a plausible, unconfirmed match to the Grgo Borić already on record as a father at house 126. See the Kalanj page. --- 30 September 2026, images 240 to 242. The register turned over into 1894. A fourth child, Ove Borić, born to the Tome Borić x Jela Borić household at house 141 (siblings Juraj 1888, Petar 1890, Matija 1892 already on record) — the first of the four births with no discrepancy in the mother's name. See the Kalanj page. --- 30 September 2026, images 243 to 248. A new Borić household at a new address: Luce Borić, born to Lojze Borić at Kamenica house 83 (distinct from the house-24 puzzle). A Borić witness, given name uncertain, at a Jelić baptism. See the Kalanj page. --- 30 September 2026, images 249 to 251. A Borić witness who may be the very first Borić this archive ever found in this parish, four years on from his own marriage. A sixth woman named Marija Borić married out of the surname, into Krnarić this time. See the Kalanj page. --- 30 September 2026, images 252 to 254. A second unmarried Borić mother in this file: Kate Borić, of Kamenica house 3, whose daughter Ove takes her surname - the same pattern already seen once before at a different Kamenica house. See the Kalanj page. --- 30 September 2026, images 255 to 258. A fifth woman named Luce Borić married out of the surname, into Bišanić this time - an eighth Borić woman married out overall. See the Kalanj page. --- 30 September 2026, images 259 to 266. A third Niko Borić household: Marko, born September 1894 at Brinje house 148, distinct from the two other Niko Borić households already on record. See the Kalanj page. --- 30 September 2026, image 267. A plausible return: Manole Borić, already on record at Kamenica house 24 as a mother by 1891, stands witness at a Kamenica baptism three years later - the same woman, probably, though unconfirmed. See the Kalanj page. --- 30 September 2026, images 273 to 278. The house-126 Grgo Borić household confirmed: Ana, born Christmas Day 1894, a sibling of the Stjepan already on record - the same parents twice over, a real household. See the Kalanj page. --- 30 September 2026, image 279. The register turns into 1895. A first Borić name at Lučane as a witness, alongside the Borić woman already married out there in 1893. See the Kalanj page. --- 30 September 2026, image 280. An eighth Marija Borić married out, into Hofjenc a second time - the same family a Marija Borić already married into at house 142. See the Kalanj page. --- 30 September 2026, images 281 to 286. A new puzzle at the house-141 Borić-marries-Borić address: Marica, born March 1895, names her parents as Ive and Matija Borić, not the Tome and Jela already on record there for four earlier children. See the Kalanj page. --- 30 September 2026, images 287 to 295. A plausible, unconfirmed link between two Kamenica Borić households: a mother's name repeated exactly, a father's name and house number that don't match. See the Kalanj page. --- 30 September 2026, images 296 to 299. A new Borić given name, Stjepan, stands witness at Kamenica. See the Kalanj page. --- 30 September 2026, images 300 to 318. No new finds - 19 clean images. --- 30 September 2026, images 319 to 327. A ninth Marija Borić, as a witness at Kamenica. See the Kalanj page. --- 30 September 2026, images 328 to 335. A second sighting of a Borić man witnessing a Milaković baptism - plausibly the same recurring witness as image 239, though not confirmed. See the Kalanj page. --- 30 September 2026, images 336 to 338. A plausible return of the Kamenica house-83 father: Alojz Borić (the full form of Lojze) witnesses a Škorup baptism - the same Škorup surname already tied to his household. See the Kalanj page. --- 30 September 2026, images 339 to 344. A second confirmed child for the very first Kalanj household: a son named Niko, after his father, born 1896 - a sibling of Anica Kalanj. See the Kalanj page. --- 30 September 2026, image 345. The first Perković woman found in this parish: Kate Perković, married into Hofjenc/Hofjevac, the same family several Borić women have also married into. See the Kalanj page. --- 30 September 2026, images 346 to 348. A fourth Kate Borić, married into Domić at Lučane. See the Kalanj page. --- 30 September 2026, images 349 to 350. A second child confirms the Bišanić-Borić household: Ivan, born September 1896, a sibling of Franjo (1894) - same house, same parents. See the Kalanj page. --- 30 September 2026, images 351 to 356. A second child confirms the Sertić-Borić household at Kamenica: Niko, born November 1896, a sibling of Ivan (1893). See the Kalanj page. --- 1 October 2026, images 357 to 369. The register turns into 1897. The Krnarić-Borić household confirmed with a second child, Tome, three years after the first. See the Kalanj page. --- 1 October 2026, image 370. A third sighting of the recurring Borić-Milaković witness pairing, at the very same house 74 as the first, in 1893. See the Kalanj page. --- 1 October 2026, images 371 to 381. A new Borić given name, Viol, survives only as a witness line from an incompletely-photographed entry. See the Kalanj page. --- 1 October 2026, images 382 to 387. A new Borić given name, Roza, as a witness at Brinje. See the Kalanj page. --- 1 October 2026, images 388 to 390. A third Jozo/Joso Borić witness, at a Škorup baptism - the same surname already tied to a plausible Borić sighting elsewhere. See the Kalanj page. --- 1 October 2026 — THE 1893-1897 ROĐENI ITEM IS FINISHED. Image 400 is the film's own end card. This one item alone added dozens of new Borić individuals, the first Perković woman found in this parish, and two more children confirming the Kalanj household this archive found first. A third item begins right after, its numbering continuing rather than resetting - probably the same book running past 1897, not yet opened. See the Kalanj page. --- 1 October 2026, images 401 to 406. A third item confirmed as a continuation of the same register past 1897 (no fresh title, same numbering). A new Niko Borić witness at Brinje house 154. See the Kalanj page. A third child (Marica) for the house-142 Niko Boric household turned up at image 427, born 23 July 1898. A second Sertic-Boric couple turned up at Kamenica house 28 (image 431); a Boricevic near-miss and a second Jure Krnaric are flagged but not merged. Alojz Boric confirmed as a household head at Kamenica house 23 (image 449), the same house already holding Tome Boric - three Boric men and a recurring Mesic wife, not yet resolved. Session paused at image 450 to check in with David; a new Kalanj entry (Jure Kalanj x Vucetic Manole) turned up at house 231, the same house as the first Kalanj household, kept separate. Resumed 5 October: a fifth child at the house-141 Boric household and a second Kalanj birth at house 231 (father Mile, mother Vucetic Marija) found on images 450-451. Images 452-454: third children for the Krnaric-Boric household (house 208) and the Viol Kalanj household. Images 455-462 (August 1899): a third Boric father at Kamenica house 23 (Stipe, wife Sertic Luce), two Kate Boric mothers married to different Sertic men within a week, Lojze Boric a witness at Kamenica 5. Images 463-472: a second child for the Tome Boric x Radotic Marija household (Kamenica 24); NEEDS A PERSON: the Kalauj near-miss may in fact be Kalanj (see addendum, images 463-472). Images 475-481: Boric Marija (wife of Mesic Stipo, Kamenica 20), Boric Luce (wife of Sertic Pave, Brinje 289), Boric Milka as witness twice, and a Kalanj Jerko household at house 188. Image 483: Kalanj Stipe at house 249 (Mijo Kalanj x Krnaric Manda) - house 249 also held the Kalauj Jure household of 1892, which strengthens the case that Kalauj is Kalanj; needs a person. Walked once 5 October 2026; reopened the same day. One question needs a person, not more walking: whether the Kalauj near-miss family is in fact Kalanj (house 249 held both); and a second look at the near-miss witness decisions made at fit zoom after image 450. 5 October: KALAUJ IS KALANJ (images 157, 171, 259, 274); seven unrecorded Kalanj entries found and added. Remaining: 305-318 and 351-355, then a retest of the other near-miss surnames. 5 October 2026, later: the near-miss retest changed the answer. Re-reading 305-318 and 351-355 at readable zoom, and the images where the log named the near-miss families, found ten people the first pass had filed as 'no new finds': Kalanj Luce 1889, Kalanj Jure 1896 and a Kalanj mother (Mesić Manole 1896); Borić mothers in the Karakaš, Petrinić and Sertić births; a sixth Tome Borić child, Kate 1896, whose margin says she died a Perković wife in 1917; and two Perković households. The 'Perhonić' near-miss is Perković: the hand writes v like n, and the first pass read one letterform two ways, so every earlier Perhonić sighting is reopened. What is not done: the rest of the births volume and the marriage book were read once at fit zoom and are not proven exhaustive for Kalanj, Borić or Perković; a readable-zoom re-read of the whole of images 1-489 is the next step, and the Perkić, Perhović, Horić, Đorić and Verković near-misses were not retested. Four more the same day from the near-miss images (14, 17, 126, 416): a Perković mother where the father's own Vuković is written 'Vukonić' on the same line, a second Perković wife in the Hofjevac family, a Borić child of 1888 at Kamenica and a Borić wife of Frane Vranić. Still to re-read at readable zoom: 134, 209, 219-222, 242, 357-368, 388-389, 417-420, 453-454, 465-466, 475-477. The full re-read is under way (5 October 2026): images 10-77 of the births volume re-read tile by tile at readable scale; 46 people the first pass missed, among them three Kalanj households at house 188, Borić households at Kamenica 23 and house 141, and four Perković women married out in 1889. Images 78-489 and the marriage book are still to do. Re-read now to image 143 (5 October 2026), 79 people the first pass missed in all, and several first-pass errors corrected, among them a house number (118 for 188), a month, a given name and three misread households. One row of this archive's own was WITHDRAWN: the Borić household at house 125 is in fact Benić, a different surname. Images 144-489 and the marriage book remain. Item 1 of the births volume (images 10-194) is now fully re-read at readable scale (5 October 2026): 305 people on record, 26 more from images 144-194, one dating conflict flagged (Ane Kalanj 1892, house 249, against what the first pass dated as Mijo Kalanj's 1893 marriage; resolved in the marriage book, which dates it 15 February 1892). Items 2 and 3 (images 198-489) and the marriage book remain. Second births book re-read to image 276 (5 October 2026): 331 people on record; the first pass's own months were wrong in five rows (January read as February or August, November as October or August), a Javor child with a Borić mother had been filed as a mere witness, and the Lučane birth of Marinčić Manole has a Kalanj mother the first pass read as Šakić. Images 277-489 and the marriage book remain. Second births book re-read to image 336 (5 October 2026): 381 people on record to the end of item 2 (image 400); item 2 is fully re-read. New: twin Kalanj boys Ivan and Viol (22 February 1896, house 188), a Kalanj Tome x Mesić Marija child, a Borić daughter born to Ive Borić x Marica Vuković who died in Zagreb in 1967, and an open contradiction over Javor Ivan's wife (Borić Ane, Borić Kate or Župančić Kate). Item 3 (images 401-489) and the marriage book remain. New in the last stretch: Kalanj Marko (house 249, 1897), Kalanj Jure (house 231, 1897, informant Borić Lojze), and three Javor households that cannot yet be told apart. Third births book re-read at readable zoom (images 404-489, 5 October 2026): 414 people on record, 33 added from item 3. Confirmed at close zoom: Jerko Kalanj x Krnarić Ruže (house 188) and Mijo Kalanj x Krnarić Manole (house 249, November 1899). New: a Perković mother (Jurajić Kate, house 286, 1899); Mesić Stipo's wife corrected from Marija to Matija; Perkić (Perkić Mijo, mothers at Vranić 265 and Javor 70) kept distinct from Perković; Javor households now at least four (houses 70 twice, 82/83, 83, 92). Only the marriage book (images 493-690) remains at readable zoom. Marriage book re-read to page 159 (images 493-572, 6 October 2026): 432 people on record. Settled: the two Javor Ivans of house 83 are father and son; the Mijo Kalanj wedding is 15 February 1892 (not 1893), so the Ane Kalanj 'contradiction' was a misread digit; Mijo and Jure Kalanj of house 249 are sons of one Kalanj Lju- and Ruže; the Perkonić/Perhonić of the marriage book is Perković (a bride of 1891 and a witness in 1889 and 1892). Marriage pages 160-end (images 573-690) remain. The whole Brinje volume is now re-read at readable zoom (6 October 2026): births images 10-489 and marriages 493-688, 461 people on record. Marriage book settled the two Javor Ivans of house 83, the Zoričić Pave wives (Perković Kate then Borić Kate), the Perlić-not-Perkić question, the Mijo Kalanj wedding date and house 126's two Borić couples. Next: the Umrli (deaths) volume, which names parents and survivors, is the likelier place to link the Brinje households to each other or to Klenovica. The Umrli (deaths) volume is read to its end (6 October 2026): DGS 005498263, waypoint 9R2D-W32, images 9-337 (pages 2-658; image 338 is the END OF ITEM card). 48 deaths are on record, 23 Kalanj (two written Galanj) and 25 Borić, and no Perković death anywhere. The volume was first read as whole spreads and then RE-READ PAGE BY PAGE at readable zoom (both pages of every image, images 9-337); the re-read reproduced all 47 first-pass entries and added one, so the count is now a page-by-page count and no longer a floor. It documents what the births could not: Janja Kalanj, widow of house 231 (d. 4 July 1889), is the mother named on the 1888 marriage of Jure Kalanj, son of Vid and Janja; the 1896 twin boys of Viol Kalanj and Kate Zoričić died on 25 and 26 May 1897; Ane Kalanj, 17, of house 231 (d. 7 August 1892) is the sister of the bride Marija Kalanj of January 1892; Mijo Kalanj of house 249 was probably widowed by Bare on 8 October 1891, four months before his second marriage; two entries written Galanj at house 151 probably belong to the Marko Kalanj household and to the Mijo and Manole of the 1893 Kalanj bride (probable, not proven); Petar Kalanj, 3, who died on 18 October 1894 (house read 181) is almost certainly the Petar baptised in November 1891 to Tome Kalanj and Mesić Marija of house 188 (the one entry the re-read added); and Viktorija, wife of Niko Kalanj of house 188 (d. 4 December 1889), conflicts with a Niko Kalanj of house 231 who married on 16 February 1890 as a bachelor, a conflict left open. Emigrants of 1909-1910 against the registers (6 October 2026): Stipe Kalanj (b. 1885) predates the books and the 1899 Brinje Stipe is far too young; Mile Kalanj (b. 1889, age 20 in 1909) has one candidate, the Mile born 21 October 1888 at house 188 (son of Tome Kalanj and Marija), not among the deaths to 1899, with the manifest's uncle Jure Božić unexplained; the cousin Juro Borić has one candidate, Juraj Borić b. 3 February 1888 at house 141, also alive through 1899. Leads, not matches. Still not established: any link from a Brinje Kalanj, Perković or Borić household to Josephus Kalanj of Klenovica or to anyone in the tree; no death entry names Klenovica, Senj or America. Brinje's filmed 1888-1899 volumes are now exhausted for these three surnames at the level described above; if this line is pursued, the places to look are the Gospić letter's Kalanj x Perković marriage and the Klenovica registers.

Found by the missing-row gate on 15 September 2026, and it is the row that proved the gate should have no exemptions — it was written off as «a plan» when it is a list of documents sitting in a book this archive has already opened. «Ana Shambul Marić» is a name, two tree links and nothing else. Her father is a surname with no given name; her mother is not in the tree at all. The name is probably not her name. She is carried as *«Ana Shambul Marić»* and elsewhere *«rođen Shumbul Marić»*, and this archive has met that shape before — the nadimak written beside the surname, *Antich Papa*, *Jellichich dictis Jurešina*. If Šumbul is the house name then the surname to search is Marić, and the by-name is the only thing separating this family from every other Marić in Lika. *A hypothesis, deliberately kept out of the by-names data, which holds only phrases read off a page* — and note that row 41 has since widened the connector list, which may bear on it. What is reachable, and it is cheap. The couple's children are Lucca (1853), Marija (about 15 September 1858), Jacobus and Tereza (1867). Otočac births are filmed 1834–1858, so Lucca's and Marija's baptisms are both inside the book — the same book as row 27, which is index-led. Do them together. --- The hypothesis is now supported from inside the tree, 15 September 2026, and it did not need a book. Written up in [notes/ana-sumbul-maric.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/ana-sumbul-maric.md); four things the GEDCOM says on its own. * «Shambul» is in the SURNAME field and the given name is empty — the record is literally `1 NAME /Shambul Marić/`. So this was never *a man called Shambul whose surname was Marić*. The whole string arrived as a surname. * His other two children are plain Marić. Mathias and Joannes, same father, same family record, no second word. A surname does not attach to one child of three. A by-name does — it follows a household, a house or a man, and the priest writes it when he needs to tell one Marić from another. * The two sources disagree, and neither is a document. A MyHeritage Smart Match (QUAY 3) whose URL reads *shumbul*, and a Geni world tree (QUAY 4) reading *shambul*. Šumbul in one, Shambul in the other, and both are other people's trees. * Her married name is recorded separately — `_MARNM Žubrinić` — so the tree already keeps apart the name she was born with and the name she was buried under, which is the distinction the Senj stones forced on the grave matcher the same day. So the surname to search is Marić, with Šumbul as the discriminator — the way *Papa* separates one Papić household at Selce and *Smojver* one Vukić at Karlobag. It stays out of `sources/bynames.psv`, which holds only phrases read off a page; this one was read off a URL, and the file it is in says so. What is left is the book, and it is the one row 27 wants. Otočac births are filmed 1834–1858 and Lucca's (1853) and Marija's (about 15 September 1858) baptisms are both inside it. Two entries would give her father's given name, her mother's existence, and whether the priest wrote Šumbul beside Marić. --- 20 September 2026 — searched the moment FamilySearch came back, and the answer is that this row is out of date about its own person. The Otočac index gives «Maria Zubrinić, 15 September 1858, of Nicolaus Zubrinić and Anna Marić» — ahnentafel 10 and 11 named together. Against a row that opens *«Ana Shambul Marić is a name, two tree links and nothing else»*, that looked like the row falling. It is the same entry `sources/readings.psv` has held since 14 September, and the reading already here is better than the index, because it carries the household the index drops: *«Rustica e Šumećica N° 2»*. And there is a second, the Otočac baptism of 3 July 1856, same house, naming her «Anna nata Štrumbulko» — the by-name attested in a register, which is what [the plan note](https://github.com/daviddef/TheBlazevic/blob/main/notes/ana-maric-plan.md) had as a hypothesis. So she was never empty, and this row's opening sentence has been wrong since the 14th. Recorded rather than quietly fixed, because the archive announced a rediscovery once before with the Karlobag index and the habit is worth catching every time. What the index actually adds is one citation and two negatives. * The 1858 reading had no ARK. It has one: `1:1:QKMK-3MZV`. * Her own baptism is not indexed, and the reason is coverage. A control search returns three Marić entries at Otočac in total — 1844, 1844 and 1852 — and she was born about 1836. The film is still the only route to her own entry, and to ahnentafel 22's given name, which no baptism of her child will ever supply. * «Shambul» is a real surname 250 km away. Every *Shambula* in the index is at Donji Mosti or Zrinski Topolovac in Bjelovar-Križevci, or Bartolovec near Varaždin — Hungarian country, nothing to do with the Lika coast. Not one is in Lika-Senj. That is negative evidence for the *nadimak* reading, and a warning: a searcher who takes the surname at face value lands among real people in the wrong county. --- 20 September 2026 — the by-name hypothesis now has evidence from OUTSIDE the file it is trying to correct, and it is a counting argument. Everything supporting «Šumbul is a *nadimak*» was, until today, internal to the GEDCOM: the string sits in the surname field, only one of three siblings carries it, the two sources disagree about its spelling. Good reasoning about a file. Searched across the whole Croatian record index: | form | exact hits in all of Croatia | |---|---| | Štrumbulko | 1 — and it is ours | | Shambul, the tree's own spelling | 0 | | Sumbulko, Strumbul, Sumbulic | 0 | | Sumbul | 7 — all Dalmatian, all with Italian given names | Marić, for contrast, is ordinary at Otočac — seven in a single page for births 1800–1880. A surname leaves a trail of cousins; this one leaves a single entry. And the seven *Sumbul* records are three hundred kilometres away in the wrong language. In the index, «Shambul» is not a surname anywhere. And the index shows the alternation itself, which the film had already shown: 3 July 1856 Stephanus Bubrinić · Nicolaus Bubrinić × Anna ŠTRUMBULKO · QKMK-SK94 15 September 1858 Maria Zubrinić · Nicolaus Zubrinić × Anna MARIĆ · QKMK-3MZV Same father's given name, same parish, two years apart, the mother's family name different in the two entries. This archive already held both — `readings.psv` has carried them since the Otočac sitting and annotates the 1856 row «the by-name attested». What is new is that it is visible in the index, in text, without the film. «Bubrinić» is Žubrinić — a third surname this index has mangled on this line in one day, beside «Piberich» and «Pillopich» for Pilipić. The query that returned both entries named no surname at all. What it does not settle. Ahnentafel 22 is still a surname with no given name and ahnentafel 23 is still absent from the tree. The marriage would name both, and it is not in the index — searched by surname pair, by spouse given names, with and without the parish, 1845–1865. Nor is Lucca's 1853 baptism. So the remaining work is still the book, exactly as this row said. It stays out of `sources/bynames.psv`, which holds only phrases read off a page. [notes/strumbulko-occurs-once-in-croatia.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/strumbulko-occurs-once-in-croatia.md) --- **28 September 2026 — the *Šumbul* hypothesis tested against the tree's own data, and it holds up as far as data can take it. 1. The vowel is unstable, which is what a heard name looks like. Her two source URLs, from two different sites, give «rođen SHUMBUL Marić» and «born SHAMBUL Marić»**. A written surname does not move its vowel between two records of the same person. A *nadimak* repeated aloud does. 2. MARIĆ EXISTS IN THIS TREE ON ITS OWN AND ŠUMBUL DOES NOT. Searched across all 1,833 published people: nine carry a form of Marić — Anna Marich, Joannes Marić, Maria Marić, Mathias Marić and the rest — and exactly two carry Šumbul, *and they are Ana and her father*. So Marić is a surname in this family's world and Šumbul is a label on one household. That is the shape of a by-name, and it is the first evidence for it that is not an analogy. 3. Her father is a surname with no given name, which is the A–Z index shape row 15 is about — so the node carrying «Šumbul Marić» is itself a sorting artefact, not a person anybody recorded. WHAT THIS DOES AND DOES NOT SETTLE. It does not prove Šumbul is a by-name; nine Marić with no places and no dates cannot prove anything on their own, and this archive's whole documented by-name evidence is Karlobag, Latin and 1776–1791 — *Smojver dictus Vukčić*, *Locatelli vulgò Žešić* — four hundred kilometres and a century from Otočac in the 1830s. The pattern is real; its application here is an inference. WHAT IT CHANGES IS THE SEARCH, and that is the point of the row. Until now there was no way to look for her. Search MARIĆ at Otočac, not Šumbul Marić — one surname, one parish, a woman marrying a Žubrinić about 1830. That is a query a register can answer, and «Ana Shambul Marić» never was.

A third kind of holder, and nobody has tried it. Not a state archive and not a parish: the Archdiocesan Office at Rijeka. vlč. Jakov Filipović, who administers Sv. Jakov Krmpote from Ledenice, says the registers of every parish he administers went to the Riječka Nadbiskupija, and that as far as he knows the Archdiocese passed them to the State Archive. What makes it worth a letter rather than a shrug. The State Archive at Rijeka answered on the same day that it does not do research of this complexity — and never said what it holds (row 10). The priest says the books are in its care. So one question is now asked from two directions and answered from neither: *does the Rijeka baptismal register covering 1 March 1892, and the Sv. Jakov Krmpote register before 1879, survive, and who holds them?* That is a holdings question, not a research request — which matters, because a holdings question is exactly what Rijeka's refusal did not cover, and exactly what an archdiocesan chancery can answer from its own transfer records. Owner is David because sending is his call. The draft can be written here on request; the three asks are the ones already refined in `requests/sent/EMAIL-zupa-krmpote.txt` and `EMAIL-rijeka.txt`, which now need only re-addressing and a covering paragraph explaining who sent us. --- 21 September 2026 — DRAFTED, and waiting only on David to send it. This row said «the draft can be written here on request». It is written: `requests/EMAIL-nadbiskupija.txt`, with the brief beside it at `requests/rijeka-nadbiskupija.md`. The framing is the work, not the asks. The asks were already refined in the two sent letters. What is new is the reframing from *please find this entry* to «what did you hand over, and to whom» — plus a fourth ask that would settle the other three at once: the primopredajni zapisnik, the inventory of what the Archdiocese passed to the State Archive for the Crikvenica deanery. An archive that will not search on a description will often produce a volume that is named, and this archive does not yet know the names. Four asks, every one a proved film-gap: Sv. Jakov Krmpote baptisms before 1879 (catalogued 1815–1896, film begins 1879, checked leaf by leaf); any Krmpote marriage register (none filmed at all); Rijeka baptisms covering 1 March 1892 (a gap by image *and* by Kazalo); and the transfer inventory. Two things deliberately left out. The Senj deaths and Milka's 1971 entry — those are civil and state-archive business, and a letter that asks for everything gets answered about nothing. And the address: it must be confirmed from the archdiocese's own directory before sending, and the draft says so rather than inventing one. Owner stays David, because sending is still his call and nothing here changes that.

Done, 15 September 2026. Every `p` row in `sources/graves/senj-2024-10-09.psv` was opened again at full resolution. 108 of 194 rows are now `r`, up from about 60. The method that made it practical: crop to the inscription and resample to about 1,800 px on the long side — the reader's own display resolution. Enlarging past that is thrown away in the downsample, which is why the first attempts at 4–8× full-frame gained nothing. Four readings were wrong and every one was a date off a contact sheet — Mile Prpić 1985 → 1982, Karlo Papić-Glavičić 1928 → 1922, Budimir Boras 1950 → 1954, and «Pavo» Boras who is *Fabo*. Two more had no readable death year at all; those years ran off the edge of the frame and were guesses, and they are `19??` now. Nothing in the ancestor block moved — all twelve direct-line and sibling rows were read at 3–8× the first time and every one was confirmed. The errors fell exactly where the confidence letter said they would, which is the case for keeping it. And the pass found a stone the first sweep missed entirely. Enlarging IMG_8159 for the Gržin inscription brought up the grave behind it: «*IVICA ŽUBRINIĆ 1887–1934 · FANA ŽUBRINIĆ 1890–1974 · MARIJAN ŽUBRINIĆ 1919–1996*». Žubrinić is ahnentafel 5's maiden surname. A Žubrinić grave in the town she married into and is buried in is a thread nobody has pulled. What is left is not another pass at the same files. 61 rows stay `p` and 25 stay `?` because the photographs do not contain the digits — steep angles run the line out of frame, flowers stand on the dates. Those rows, and IMG_8071's fallen «MARIJAN …» under the reeds, want ten minutes in the cemetery.

The re-grade of row 45 ended by naming its own successor. Four of the five «reachable» walls want one book: Karlobag baptisms, 1691–1861, filmed as DGS 5481648, and nobody has opened it for any of them. | Ahn | Who | Baptism to look for | |---|---|---| | 104 | Jacob Pilipić, *captain of the port* | about 1741 | | 108 | Georgi Uroda | about 1750 | | 212 | Josiph Smojver Vukić | 1731 | | 220 | Nicolaus Gerkacs | 1717 | All four are inside the filmed range, and a baptism is the one document that names both parents — which is exactly what a wall is short of. Four walls could fall in one sitting. It is not a cold start. The film is the archive's best-mapped: row 23 measured the deaths volume's folio anchors and its KAZALO letter positions, and `sources/familysearch/film-005481649-karlobag.json` holds the ARKs. The baptisms are a different film — 5481648 — and [the Karlobag note](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-films.md) warns that its film number does not resolve to one book: the 1771–1804 baptisms sit under waypoint `9RK1-3Y9:391644801,391849401,391849402`. Check the waypoint before trusting the film number, and expect the 1691–1770 volume to be a separate one again. And look for the by-name. Two of the four are already known under one — *Smojver vulgò Vukić* — and the Karlobag indexes file households under the *nadimak* with a «vide». --- Why the film number does not open it, settled 15 September 2026. Not a quirk and not a bad catalogue entry. **DGS 5481648 is *items 4 to 5* of physical roll 1419107, and items 1 to 3 of that roll are the parish of KAPTOL — `koha:655210`, three hundred kilometres inland. `/search/film/005481648` opens the roll at item 1, so it lands in Kaptol every time. Thirty-one of this archive's eighty-six films are in the same position; row 42 has the whole table. The route in is the waypoint, `9RK1-3Y9:391644801,391849401,391849402` — 327 images, «Births (Rođeni) 1691-1770 · Births 1771-1804 · Birth Index (Kazalo rođeni)». Known ARKs inside it: image 217 = `3:1:3QS7-L99C-L998-5` (1781), image 250 = `3:1:3QS7-L99C-L99J-1`, image 310 = `3:1:3QS7-L99C-L99Q-L`. Image 310 is inside the Kazalo — a narrow-column index page, checked at contact-sheet scale — so the index is near the end where it should be, and the four baptisms are lookups rather than a walk. A fifth wall joined the queue the same day. Row 56 put ahnentafel 426, Simoni Pekass, at Smiljan — a real parish whose films begin in 1834, which is a century too late for him. His daughter Antonia married Josiph Smojver Vukić at Karlobag on 20 November 1752, so her baptism is in this same volume and will name him. Five walls, one book. What stops it is not FamilySearch. The Browser pane is hidden in the app, so screenshots cannot composite and a register page cannot be read at full resolution — see row 21. --- 17 September 2026: this row was wrong about its own book.** It has stood for days as *«four walls want the same unopened book»*, and a session was spent on how to get a legible page out of a hidden browser pane. The book does not have to be opened. It is INDEXED, and the index is text. And this archive already knew — which is the part worth writing down. `found.psv` holds six children of ahnentafel 108 — Dominicus 1764, Petrus Jacobus 1767, Heva Joanna 1769, Maria Catharina 1772, Anton Joanis 1780, Joannis 1784 — every one citing «Karlobag births, indexed». A previous sitting worked the Uroda household entirely out of the record index, and found.psv also holds Antonius Gergacs, son of ahnentafel 220. This row was written anyway, naming four other walls in the same parish as waiting on an unopened film. The catalogue confirmed it independently on the 16th — `fs_indexed: "Y"` — read, quoted, filed. So this is not a discovery. It is a tool the archive had in its hands and applied to one household out of five. **A search of *Croatia, Church Books* restricted to Karlobag, 1690–1780, returns baptisms with BOTH PARENTS as searchable text: Roxanich 1704, 1705, 1707; Bevardich 1742 and 1752; Bovari 1749. That is the volume, and the four baptisms are lookups in a search box rather than a walk through 327 images. And it produced a household within minutes. Ahnentafel 426 Simoni Pekass × 427 Mara Pekass have no dates at all in this archive; their only anchor is that their daughter Antonia married at Karlobag on 20 November 1752. The index has a Šimun Pekasa at Karlobag: Margarita 28 Oct 1719 (father Simon), Katarina 19 Jun 1734 (Simuna and Stoja), Perina 29 Jun 1738 (Simuna and MARE). A candidate, not a proof, and the difference matters. What agrees is a given name, a wife's given name, a parish and a window. Antonia's own baptism has not been found** — *Antonia Pekas* at Karlobag returns ten records and none is hers. And there are two wives: Stoja in 1734, Mare in 1738. Which one Antonia belongs to decides whether ahnentafel 427 is Mara at all. It also contradicts row 56, written yesterday, which put Simoni Pekass at Smiljan on the tree's own place string and ruled him unreachable because Smiljan's films begin in 1834. Both can be true — a man born at Smiljan can marry into Karlobag — but the reachability verdict changes, because the Karlobag index reaches 1704. And the image route is not wasted, it is demoted. A chunked, hash-verified transfer out of a hidden pane now works — about twelve round trips per page, SHA-256 matching end to end. Images are for what the index does not carry: the house number, the *Conditio* column, the by-name beside the surname. Those are still worth the twelve. --- One new person, read out of the index and into `found.psv`. Margaritha Smojver, baptised 9 July 1764 at Karlobag, entry 270 — *«Entry for Margaritham Smoiver and Josephi Smoiver»*, mother Antoniæ, ark `1:1:QKMX-YLY7`. That is a daughter of ahnentafel 212 × 213 the tree does not have: born twelve years after their marriage of 20 November 1752 and twelve years before Josiph's death in 1776. The other two walls resisted, and the reason is spelling. *Nicolaus Gerkacs* (ahn 220, b.1717) returns nothing at Karlobag in range — the index renders the surname *Gergacs, Gergocz, Geraecs, Gergich*, and the Karlobag Gergacs households it does hold are Antonio × Francisca and Michael × Lucia, 1777–1789, a generation too late. *Jacob Pilipić* (ahn 104, ~1741) is worse: at Karlobag the index writes him Pidepics and Pilleppics. The index inherits every difficulty of the hand it was made from — it removes the need to find the page, not the need to know how a name was spelled. --- Searching by FATHER separates two Šimuns at Karlobag, a lifetime apart. The early one — Joseip 1 Mar 1705, Margarita 28 Oct 1719, Katarina 19 Jun 1734 (mother STOJA), Perina 29 Jun 1738 (mother MARE) — is the household that fits ahnentafel 426 × 427, because Mare is Mara. The late one is *Simone Pekkas × Maria*, 1783–1795. He spans thirty-three years of baptisms and two wives that are not the same woman. Either Šimun married twice in a long life or there are a father and a son of one name. Nothing found decides it. And the Antonia the index offers is the wrong Antonia. *Antonia Pehasa*, baptised 18 January 1742, parents Osipa Pehasa and Monde — she cannot be ahnentafel 213, who married in 1752 and bore Vicentius in 1754: that would make her ten at the wedding and twelve at the birth. She is in `found.psv` specifically to be ruled out, because the next reader will meet her first. Her father *Osipa* is Josip — and a Joseip Pekasu was baptised to Šimun in 1705, which would make her Šimun's granddaughter. So ahnentafel 213 was born between 1719 and 1734, in the gap between Margarita and Katarina, and her baptism is not in the index under any spelling tried. The index holds four of Šimun's children and evidently not all of them. For that one entry the image is still the answer, and the twelve round trips are what it costs. --- 20 September 2026, with FamilySearch signed in again — and the first thing the index found was a fault in this archive, not in the register. Ahnentafel 428 and 429 were placed at Karlobag three days ago «on the strength of their grandchild's death — the tree's own words about the tree's own people, not a document». There was a document. `sources/readings.psv` had held it since 10 September: > `caetano-bacchi|baptism|1779|…|Karlobag, 12 January 1779 — "ex Angelo et Catharina Bacchi"` Their names are in the citation. The reading was filed under their son and never reached them, so `walls.py` went on printing *«this wall is reachable and has not been opened»* about two people whose names sat in the archive's own file. Confirmed in the index the same day as `1:1:QKMX-BWCR`, where the surname is written BECHI. The rule existed and no tool had it. Question 9b has said since the 14th that *«a baptism counts as a reading for the child AND for both parents, because it names all three»*. That was being done by hand — sixteen times — and missed thirty. `tools/readingslib.py` holds it now and both tools that grade a person read the file the same way: 20 parents credited with a document that names them. A derived credit is marked as derived and carries the child it came through, because *«a register names you»* and *«a register is about you»* are different claims. --- THE INDEX NAMES THE WRONG MOTHER FOR A DIRECT ANCESTOR, and that is the other half of this row. Searched at Karlobag, the index gives ahnentafel 106's baptism — 18 January 1754 — as *«Osipa Smoiver × Marie»*, with the same father's children by *Antonia* in 1752 and 1764 either side of it. Read from the search box alone that is a conflict on the spine. It is not. The entry was read at full resolution on 10 September and the book is kept in Croatian: *«Isti dan kerstih vicencia Fabiana sina Ospa y Antonie Smoiver»* — son of Osip and Antonia. The register names Antonia, the tree is right, and the index is simply wrong. This row's whole finding was that the book *is indexed* and need not be opened. That still holds. The correction is that an index finds the page and the page settles the fact — the spelling failures it was already known for only hide a person, whereas this one names the wrong woman, legibly, on the direct line. [notes/karlobag-index-is-wrong-about-ahn-106.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-index-is-wrong-about-ahn-106.md) --- Three more things out of the same sweep. * A candidate for ahnentafel 104 at last — «Jakomo Pilipicha, 30 November 1739», father «Jakoma Pilipicha». The only Pilipić baptism in the index before 1784, and the spelling that finds it is the Croatian Pilipicha, not the *Pidepics* and *Pilleppics* three sittings tried. Offered as a candidate: «about 1741» is itself an estimate, and the image would give the mother and the house that the index does not. * Two children the tree does not have, both now in `found.psv`: Antonü Bechi, 11 September 1780, a second son of ahnentafel 428 × 429; and Hevam Uroda, September 1773, a daughter of ahnentafel 108 × 109 — the second of that name, after Heva Joanna of 28 December 1769. * The index holds the same entry twice under two arks. Caetano's 1779 baptism is both `QKMX-BWCR` and `QKMX-BWCT`; Antonü's 1780 is `QKMX-B9R1` and `QKMX-B9TM`. Two arks, one event — and anything counting index rows as people, or looking for children born on one day, will read it as two. --- And then the query changed, and the index stopped being a way to find a person and became a way to rebuild a household. Every earlier sitting searched by person — a surname, a given name, a year — which finds somebody you already know enough about. Searching by PARENTS returns the whole household at once. Ahnentafel 110 × 111, Michael Gerkacs and Lucia Tomljanović — six children, and this archive had one. Magdalena 1787, Maria August 1789 (ahnentafel 55), a second Magdalena 1792, Catharina 1794, Bartholomæus 1797, Mathias 1801. * Ahnentafel 55's own baptism was not here. The archive held her marriage of 1807 and her death of 1834 and never the entry she was christened on. * And it names ahnentafel 111. Lucia Tomljanović was one of the twenty-four direct ancestors with no document of any kind, placed at *Montani Stanište*, which is not a parish and never was. She is named in a register six times. Twenty-four became twenty-three, and the direct line went from 46 read to 47 — through the parent-credit rule written this morning, which is the second time today it has paid on the spine. And the by-name did exactly what this row predicted. It said *«look for the by-name — the Karlobag indexes file households under the nadimak»*. They do. Ahnentafel 106 × 107's children are indexed under both names: *Maria Vukich 1776*, *Josephus Franciscus Vukich 1777*, *Anna Oliva Vukics 1785* — and *Smojver* in 1782 and 1788. A searcher who queries only Smojver finds two of the five. The proof that it is one couple is the archive's own 1781 reading: *«ex Vincentio, et Oliva Smoyver, vulgò Vukić, Conjugibus»*. Ten children into `found.psv`, and ahnentafel 55's baptism into `readings.psv` marked INDEX ONLY, because nobody has opened the page and this same indexing has been caught twice today getting a name wrong. [notes/karlobag-households-indexed.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-households-indexed.md) --- 20 September 2026 — a hypothesis this row could have been built on, tested and killed the same hour. Two surnames showed the same shape in the index at Karlobag: Pilipić silent 1739–1810, Gerkacs silent 1721–1773. That looked like a hole in the indexing sitting exactly over the generation of this row's four walls — which would have meant the film does need opening after all, for precisely the people this row says it does not. It is not a hole. A decade census of Karlobag from 1690 to 1780 returns 100 Karlobag baptisms out of every 100 fetched, in every decade. The index is dense there. The gaps were spelling, not coverage. And the fix that follows from it found a household in one query. Searching *father Jacob × mother Catharina* at Karlobag with no surname at all returned ahnentafel 52's baptism of 11 March 1768 and three brothers — filed by the index as Piberich, Pillopich and Pillepich. See row 26. So this row's thesis survives, with its reason sharpened. The book need not be opened because it is indexed — but a surname query is the wrong instrument for a book whose hand the indexer could not read. Query the given names and the parish, and let the surname fall out. Ahnentafel 220 and 110 resisted even that. *Michael, father Nicolaus, Karlobag, 1757–1767*, surname-free, returns nothing of his; ten spellings of Gerkacs return no Karlobag baptism between 1721 and 1773. Ahnentafel 221 is still absent from the tree — Michael has a father and no mother, and his baptism is the entry that would name her. --- 20 September 2026, later — the surname-free query turned on this row's own walls, and it mostly returned nothing. Which is worth writing down as nothing. The method that found ahnentafel 52 under «Piberich» — given names, no surname, a date window, the parish — was put to the four walls and to ahnentafel 105 and 213. Four clean negatives: | Ahn | Query | Result | |---|---|---| | 105 | Catharina, father Vicentius, mother Maria, 1724–1730 | 0 | | 104 | Jacob, 1735–1748 | nothing of his | | 212 | Josephus, 1728–1735 | nothing of his | | 220 | Nicolaus, 1713–1721 | nothing of his | Ahnentafel 105's own baptism of 30 March 1727 is not in the index under any name at all, which is a stronger statement than «not under this surname» and settles it: ahnentafel 210 and 211, Vicentius and Maria Luckinić, cannot be named from the index. The 1691–1770 film is the only route left to them. And the index offers Šimun Pekas a SECOND Antonia, which is also not ahnentafel 213. «Antonia Pekasovicha, 25 October 1711, father Simuna Pekasovicha», ark `1:1:QKMX-YSF2` — a sixth spelling of the surname. She is ruled out on arithmetic: ahnentafel 213 bore Margaritha in 1764 and Joanna in 1770, which this woman would have done at 53 and 59. She is in `found.psv` beside Antonia Pehasa of 1742 — the index holds two Antonias to this household, one too old and one too young, and ahnentafel 213 is neither. That narrows her. Bearing a child in 1770 puts her birth after about 1725; marrying in 1752 puts it before about 1735. 1725–1735, in the gap between Margarita 1719 and Katarina 1734 — and her baptism is not in the index. For that one entry the image is still the answer, exactly as this row said. --- 21 September 2026 — the waypoint numbering this row uses is now convertible to film images. This row records the 1691–1770 volume by waypoint `9RK1-3Y9:…` with known ARKs at waypoint images 217, 250 and 310, and warns that «the film number does not open it». Both remain true, and there is now a bridge: DGS 005481648 has 596 images. WAYPOINT IMAGE N = FILM IMAGE N + 269 Checked on two of the three ARKs this row already carries — waypoint 250 = film 519, waypoint 310 = film 579 — rather than on one. So «image 310 is inside the Kazalo» means film 579, and the Kazalo occupies about film 569–596. And the Kaptol problem this row documents is bypassed entirely. The row's finding is that `/search/film/005481648` opens the roll at item 1 and lands in Kaptol every time, so the waypoint is «the route in». There is now a better one: the film enumerates in a single call and any image is fetched by `apid` at `<das>/v2/<apid>/dist.jpg`, which does not go through the film viewer and so cannot land in the wrong item. Thirty-one of this archive's eighty-six films are in that same position — row 42's table — and this route sidesteps it for all of them. [notes/image-75-reads-vivoz-and-the-film-enumerates.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/image-75-reads-vivoz-and-the-film-enumerates.md)

All forty-seven answered, 15 September 2026, and the gate now fails on them. `check_site.py` no longer prints a count and carries on: a contradiction about a person with no page fails the build exactly as one about a person with a page does. The bargain `checkcovers` struck — print it until the hole is closed, then close it — is over. Not one of the forty-seven needed a register. The row expected the opposite: it said *do the namesake trap first, it is the one where this archive has an instrument nobody else has*. The instrument is the house number in the *Conditio* column, and it turned out to answer a different question. Naming which Maria she is needs the book. Showing that the join is wrong only needs the tree's own arithmetic, and the arithmetic was sitting there the whole time. One shape accounts for most of them. *One spouse slot holding two generations of mothers.* Vladimira Baričević, born 12 December 1907, is entered as the mother of all fourteen children of a man born 1855 — the eldest of them born 1888. The tell is that Zdenka Boras is born 12 December 1907 too: a daughter's birth date copied into the wife's record. Helena Untersinger is the same shape — five children across thirty-seven years, the first when she is fourteen and the last eleven years after she is buried. So are both Žubrinić Marias, and Catharina of Prozor 50, whose eighth child arrives nineteen years after his father dies. The second shape is a sibling recorded as a child. Marko Butković, born 1872, is the father of Antonija Butković born 1880 — who carries his surname, dies at eleven, and is his sister. Luka Katnić died in December 1755 and fathers a daughter in 1757 who carries her *mother's* surname: his widow remarried and her second family was hung on her first husband. And three were never contradictions at all. Two are hedged dates — *AFT 1926*, *BEF 1750* — doing exactly what a hedge is for, which `/impossible` already files as «not an error». The third found the only gap in the tool: Marija Šojat's outlying child is «ABT 1840», and the check misses the hedge because it reads the person it is reporting and not the person on the other end of the join. A bucket fell out of it. Joanis Grach's father is entered as *«Grah's for Investigation (not real)»*. This archive audits the tree for sorting buckets and did not have this one: the phrase list was written from the buckets that were causing trouble, so it knew *for sorting* and *to be sorted* and *it seems*, and not PLACEHOLDER, for Investigation or (not real). Five were missing, parenting thirty-six people between them, none of them published — which is why nothing on this site was ever wrong, and why nobody noticed. A detector written from the examples in front of it will always be shaped like those examples.

Both settled as places, 15 September 2026, and neither settled as a book — which row 45 was right to insist are different questions. Smiljan is its own parish. `koha:465540`, *Rimokatolička crkva. Župa Smiljan* — not a village of Gospić, which is exactly what this row asked. It is now in `parishes.psv` with its two films. It does not reach ahnentafel 426. Smiljan's births are filmed 1834–1857 only, and Simoni Pekass is a generation of about 1700: his daughter Antonia was already marrying in 1752. Gospić was added beside it for one reason — it is the only book in the district that goes back further, and only for marriages and deaths from 1789. So the way to Simoni Pekass is sideways, and it is a book this archive already wants. Antonia Pekass married Josiph Smojver Vukić at Karlobag on 20 November 1752. Her baptism is inside Karlobag Rođeni 1691–1770 and will name her father. That is row 54's volume — a fifth wall wanting the same unopened book. Montani Stanište is not a parish and the catalogue agrees. Searched for both *«Stanište»* and *«Montani»*: nothing in Croatia. A *stanište* is a herders' settlement and its events are written into a neighbour's book. Which neighbour is not established, and no row was invented to pretend otherwise — `unknown` is a legal value and that is what it carries. What is firm is the daughter: ahnentafel 111 died 19 September 1809, inside Karlobag *Umrli 1804–1861*, and that entry would name her father and her house. A tool fell over on the way. `walls.py` accepted a `why` override and then silently dropped it unless it arrived with a `place`, because it had only ever been written as the second half of *«a document places them, and here is the document»*. Both sentences above were parsed and thrown away on the first run. A wall can need a sentence without needing to move. --- Qualified on 17 September. This row ruled ahnentafel 426 unreachable because the tree places him at Smiljan and Smiljan's films begin in 1834. The Karlobag record index reaches 1704, and it holds a Šimun Pekasa fathering daughters there in 1719, 1734 and 1738, the 1738 entry naming his wife Mare — against this archive's Mara Pekass, ahnentafel 427. The sideways route this row proposed is the right one; it is the verdict «not on film» that was wrong. See [notes/karlobag-is-indexed.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/karlobag-is-indexed.md).

16 September 2026: a full month's subscription, where before there were three trial days. That changes what is worth doing, because the trial's own closing note said what it could not do: *«Perković (2,036), Borić (2,007), Papić (1,042), Prpić (943), Boras (3,445) are too common to harvest in bulk. They need a first name and a year.»* `sources/findmypast/targets-2026-09-16.psv` is the first name and the year. Generated from the archive rather than typed, so it cannot drift from the tree: every published person born 1840–1910 with no death recorded — old enough to sail, young enough to be on a post-1906 manifest. 386 people. Žubrinić 115, Kalanj 95, Prpić 54, Blažević 50, Papić 25, Šestan 23, Perpić 12, Boras 7, Vukelić 4. It is not a list of emigrants. It is the list of people this archive stops following — which is the same list until somebody looks. Not one is a direct ancestor, and that is worth saying plainly: this row buys breadth, not the line. Work it by surname, not by person. One surname read end to end beats forty name searches, and the trial proved why: `ZUBRINICH` returns 191 records and gives Australia; `ZUBRINIC` returns a different 119 and gives Lika — Šumećica, Otočac, Prozor. Anyone searching one spelling is missing the other entirely. The file carries a spellings column with every variant this archive has met. The four things the trial left unfinished, and they come first: 1. Port Pirie electoral rolls — eight of nineteen unread. This is the one already sitting in `searched.json` as pending. Rolls give an address and an occupation year by year. The archive records a trade for 16 people out of 1,245; these could outweigh that several times over. In Lika this family held offices — forest warden, master of the watch, keeper of the prison. At Port Pirie they are labourers in a lead smelter, and the trade page says so on eight men. The offices did not travel. 2. The remaining ~13 `zubrinic` manifests — 106 of 119 were read. More Prozor and Šumećica residences, more named fathers. 3. South Australian birth, marriage and death registrations for the eighteen Australians. The one already done settled a duplicate: one 1886 Port Pirie registration, for a *Kate*, and no *Katalyn* in 176 records. 4. Boras at Yarraville (1892) and Sydney (1919) — never researched, and Boras is one of only two families that reaches Hedviga without scaffolding. Two negatives already paid for, so they are not re-tested. FindMyPast holds no Croatian parish registers — so it cannot answer Brinje, and row 50 still wants the film. And there is nothing on Hedviga's 1948 emigration: no *Hedviga Blazevic* anywhere, no Defranceski in *Passenger Lists Leaving UK 1890–1960*. They went to South Africa and this service has neither end of the voyage. Method, from the trial, worth a day on its own. Only the pre-1925 Ellis manifests (`ELLIS2` transcripts) carry a last residence and the relative left behind — that residence is what turned the Kalanji from a Klenovica family into Brinje people. Ids ending `…/2` and `…/3` are relations named on a record, not passengers. --- Item 1 is done, 16 September 2026, the same day. All fifty Australia Electoral Rolls transcripts for *Zubrinich* — nineteen people, five roll years, in `sources/findmypast/electoral-rolls-2026-09-16.psv` and written up in [notes/port-pirie-electoral-rolls.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/port-pirie-electoral-rolls.md). The `searched.json` row that had been pending since the trial expired is closed. The archive's occupation count went from 8 of these people to 23, and `/work` gained a fifth stratum — *Port Pirie and Whyalla — the lead smelter* — which is now the second-largest after Senj. It should have existed already: this archive has been saying «the offices did not travel» while giving the destination nowhere to appear. Two trades and no third. Foundry — *ironmoulder*, *moulder* — and *labourer*, plus one shop assistant. Not one office, not one rank, against *Silvarum Custos*, *Vigiliarum Magister*, *Custos Carcerum* and a municipal excise controller in Lika. And the rolls do what a register does. Keith Leo and Rita Laurel move 9 Queen Street → 46 Cottell Street → 58 King Street in four years. Two marriages fall between roll years — Asta Louise comes from an Adelaide suburb to Whyalla, Una Blanche from 11 Queen Street to 30 Beach Road. Three separate Zubrinich households stand in Queen Street, at 9, 11 and 65. One caution, proved inside this set: the «household members» field is not a household. Maurice Joseph and Violet Mary share 56 Simpson Road in all three years and only the 1941 transcript lists them together. Group by address, not by that field. And the spelling lesson recurs one level down. *Austa*/*Asta Louise*, *Benedicter*/*Benedieter*, *Valentino*/*Valentine*, *Una*/*Una Blanche* — four of nineteen indexed twice inside a single record set. --- Items 2 and 3 done the same day. Item 2 — the manifests, and the spelling count. It was written here as «three and no more» and that was wrong within the hour. Zubrinich 191, Zubrinic 119, Zubrinick 10, Zubrinec 1; Subrinic, Xubrinich, Zubrinicz zero. The first test was six guesses at how a CROATIAN ear would render the name. Zubrinick is how an AMERICAN CLERK did, and no guessing from the Croatian side reaches it. The spelling that finds a family is the one written by the person who could not spell it, in the country they arrived in. The single `Zubrinec` record is the best thing on this list. *Franc Zubrinec*, b.1887, *Adriatic* 1911, residence «Sinae, Hungary» — SINAC, on the Gacka near Otočac — father Pavao. `parishes.psv` has carried Sinac since 14 September, with three films, and not one person in this tree has ever been placed there. It has one now, and Sinac births are filmed 1834–1901, so his 1887 baptism is inside the film. He sailed on the same manifest page as Anton Zubrinic of Prozor — the record ids are three apart. Also: Eva Žubrinić of Prozor sailed from FIUME on the *Ultonia* in 1907, where every other departure here is Southampton, Rotterdam or Liverpool. And Gomerčić is a second family married in after Kostelac — Eva met by a *brother-in-law* Joso Gomerčić, and Mijo Žubrinić recorded as *uncle* to the passenger Pavdo Gomerčić. Neither family is in this tree. Item 3 — the South Australian registers, and they carry the parents in the index row. Thirty-five births, eleven marriages, fourteen deaths, no transcript needed. John × Rose marry 1884 and have twelve children to 1909; five of the six labouring men in the rolls are on that list and the sixth is a grandson. The address guesses from the rolls are now on birth records — Malcolm Theodore is Edward Malcolm's son, Gwenyth Florence is Arthur Leslie's daughter. Michael Valentine lives alone in every roll because Minnie Doris died in 1932. And they corrected me twice, within hours. Thomas Leo and Una Blanche married in 1912 with nine children, not between the 1939 and 1941 rolls; Leonard Francis and Asta Louise at Adelaide in 1936, not between 1939 and 1943. Both slips came from reading the roll's «household members» field as a household — the very trap the same note had already written down, and it still caught me. Corrected in [notes/port-pirie-electoral-rolls.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/port-pirie-electoral-rolls.md). Item 4 is thin and the trial was right about it. The archive's own Boras with no death are all at Senj — Christina Maria 1862, Franciscus Paul 1872, Marija 1892 and the rest. The Australian Boras are the English-named NSW family the trial dismissed. Yarraville and Sydney are not reachable by a place filter — FindMyPast's results URL does not take one — so this needs a name-and-year pass off the target list like any other common surname. --- PERPIĆ was tried, and it is a clean negative — which changes how the rest should be done. `PRPIĆ` returns 943 and `PERPIĆ` 112, the Zubrinich/Zubrinic split exactly, so it looked like the easy one. Four Perpić passengers carry a residence and they are four different places — *Kastav*, *Matkovići*, *Karlovac*, *Zeria* — not one of them Krivi Put or Senj, and no two near each other. The other hundred are an American family at Perry, Wyoming County, New York, with a later limb at San Diego and Bergen County. That is the Kalanj lesson again, and it should now be a rule: the trial wrote *«Kalanj is a regional surname, not one household»*, and Perpić is the same. A Croatian surname on a US manifest index is a region, not a family. Harvesting one in bulk buys nothing without a place to test it against — so Šestan 519, Papić 1,042, Blažević 1,374 and Vukelić 2,424 should not be swept at all. They should be worked name and year at a time off the 386, which is what that list was built for. The sweep gave two things anyway. Joka Perpić's 1911 record spells the daughter PERPIC and her father PRPIC in the same entry — the two spellings shown to be one surname by the index's own hand. And she sailed from FIUME, the second Fiume departure in this material after Eva Žubrinić in 1907. Fiume is a route this evidence keeps showing, and it is three hours from Senj rather than a crossing of Europe. --- And the obvious escape from that rule was tested and does not exist. If a surname returns a region, then search the village instead — FindMyPast's results URL takes `keywords=`, and the form calls it *«Optional keywords»*. It looked like it worked: `keywords=smokvica` returns 13 passenger records. They are Banićević, Gabrijel, Pešelj, Radovanović, Šalečić, Šeparović, Teodorović, Tomašić — and the first transcript reads «Residence: Smokvica, DALMATIA». That is the Smokvica on Korčula, four hundred kilometres from Smokvica Krmpotska. A village name is no more unique than a surname. Then every one of this archive's villages returned nothing — *krmpot*, *krivi put*, *sumecica*, *klenovica*, *jablanac*, *sinac*, *prozor*, all 0 — including the misspellings the records themselves use: *prozov* 0, *otochetz* 0, *sumenica* 0. The decisive test. Eva Žubrinić's 1907 record gives *Residence: Prozor* and *Ship: Ultonia*. `lastname=zubrinic&keywords=prozor` → 0. `&keywords=ultonia` → 0. Her own record is not returned by her own village or her own ship. The residence is on the transcript and not in the index: readable once you have found someone, useless for finding them. So there is no way in by place, and a common surname returns a region. The only handle is a name with a year beside it. `targets-2026-09-16.psv` was built as a convenience. It turns out to be the only instrument there is, and the rest of the month should be spent working down it. --- The instrument was here all along: the NATURALISATION PETITION. A manifest gives a residence and a relative. A petition gives an exact birth date, an exact birth place, the spouse's birth date and every child's birth date — a whole household in one row. Ivan Žubrinić's also carries the alias «Or Rozić» — which a first reading took for his wife's maiden name and is not: the index attaches it to Ivan and all three children too. It is an alternate surname for the whole household, and reading it as a maiden name would have invented a marriage line that is not there. And in this set the birthplace IS indexed, where in the passenger lists it is not: `keywords=otocac` 7, `keywords=senj` 12 — against 0 for *prozor* in both. The village search works after all: on towns, in the petitions only. That is the seam the rest of the month should be worked in, and it is a better one than the 386. Two Senj-born emigrants in this archive's own surnames, neither of them recorded here. Marijan Vukelić, b.1894, born Senj, five filings 1935–1948 (also indexed *Vjkelic*) — the line Milka Papić married into. And Miro Šojat (*Sojath*), b.1910, born Senj, 1955. A trap that nearly put a false negative in the archive. Four surname-and-dataset searches came back 0 — Blažević, Prpić, Papić, Šestan in the passenger lists — and *«no Blažević emigration records exist»* was one keystroke from being written down. They were not zero. The check polled for the word *«results»*, which appears in the filter panel as *«View N results»* before the count renders. Polling for a digit followed by «results» gives Blažević 706, Prpić 632, Papić 529. A negative is a finding and has to be earned like one. The corrected Blažević probes are honest negatives: *Mile* → Dvor, *Joso* → Kuthuryevac, *Jurko* → nothing. --- Four petitions read out, and the seam is Otočac and Senj. `keywords=` gives Otočac 7 and Senj 12; Karlobag 0, Novi Vinodolski 0. The match is fuzzy — searching *klada* returns a page largely of Greek names — so a hit is a candidate and the transcript is the test. All four below were opened and read. KLADA — a village in no parish table, no gazetteer, and no person's record. Marijan Vukelić, born 27 January 1894 at *«Klada-senj»* — the coast below Sveta Jelena — wife Rosie b.1896, children Walter 1922, Catherine 1925, John 1927, Brooklyn, five filings 1935–1948. The same shape as Sinac this morning, found the same way. And Vukelić is not a stranger's name: it is the line Milka Papić married into, and this archive records no Vukelić emigration at all. Three more, all confirmed. Miro Šojat (*Sojath*), b. 5 Nov 1910 at Senj, Astoria NY 1955 — Šojat runs from Marija Šojat of Krivi Put to the Nikola Šojat on the Senj stone, and none of them has an emigration recorded. Luka Orsanich — *Orešković*, ahnentafel 21's surname — b. 15 Oct 1897 at Otočac, the Bronx 1931. And Kate Žubrinić, b. 27 Oct 1895, born at Otočac: the archive knew her husband was, it did not know she was. Two sworn records of one family, and not one child's date agrees. Ivan and Kate filed separately and both were sworn: George 7 Jan 1923 / 21 May 1922; Joseph 8 Feb / 2 Feb 1924; John 13 Nov / 8 Oct 1932. Kate's own birth year is 1905 in his 1936 filing and 1895 in his 1939 one. The most formal record this family left in America is not reliable to the month. Excellent for *who* and *where*; no better than a manifest for *when*. Written up in `sources/findmypast/naturalisations-2026-09-16.md`. --- The surname box matches EXACTLY, and that is the third form of the same lesson today. Four of this archive's surnames returned 0 in the petitions — *kalanj*, *papic*, *sojat*, *spalj* — and Miro Šojat is demonstrably in the set. `lastname=sojat` 0; `lastname=sojath` 4. A zero on a surname means «not under that spelling», never «not here». (The *keywords* box is the opposite — fuzzy enough to return Greek names for *klada*.) Zubrinich/Zubrinic/Zubrinick/Zubrinec; Perpic/Prpic spelled differently in one entry; now Sojat/Sojath. One letter, and the search is blind. Kalanj is «Kalan» in America — and the Kalan family is Slovene. `lastname=kalan` returns 20 petitions the 143-record Kalanj harvest never saw, and it looked like the find of the afternoon. Frank Kalan, b. 10 April 1873, birthplace «Krain, Austria» — Carniola, Slovenia, not Lika. *Kalan* is a Slovene surname in its own right, not a rendering of Kalanj. Recorded so nobody spends a day on it. Same for the second Vukelić household: Louis Peter Vukelić, b.1892, birthplace «Rogul», New Haven 1926 — not Klada, not Senj, no relation established. And Brinje is confirmed as an emigration source, in the clerks' own spelling. Vid Peter Karakas, b. 26 September 1920, birthplace «Brinje Plasca, Yugoslavia» — Brinje AND Plašćica in one field. The trial found seven Kalanj manifests giving *«Brinje, Plaščica, Kurjak and Velgovica»*; here is the same pairing thirty years later in a different record type. Karakaš is not one of this archive's surnames, but the form the clerks wrote is — and row 50 wants exactly that. Where the villages stand: Otočac 7, Senj 12, Brinje 1 (fuzzy); Karlobag, Novi Vinodolski, Jurjevo, Selce 0; Ledenice only a Czech *Karel Tuma*. The seam is inland and it is Otočac. The coastal parishes this archive has read most sent nobody to New York under a searchable spelling. --- All nineteen filings read: fourteen people, and thirteen of them went to New York. `sources/findmypast/senj-otocac-naturalisations.psv` is the register — twelve Senj filings and seven Otočac ones resolving to six and seven people, plus Brinje's one. Brooklyn, the Bronx, Astoria, Jackson Heights, Long Island City, Tarrytown, Nassau. Only Mike Mukulino, the oldest, born 1858, went elsewhere — to Boston. That is a third diaspora, and it touches neither of the two this archive already had. The Žubrinići of Prozor went to Port Pirie and became moulders in a lead smelter. The Kalanji of Klenovica went to Leadville, Colorado and the Michigan copper country. Senj and Otočac went to New York. And «Birth country» turns out not to be evidence. The same town, by American clerks: *Senj, Hungary* (1920) · *Senj, Austria* (1928) · *Senj, Hrvatska, Austria* (1924) · *Senj, Yugoslavia* (1928) · *Senj, Jugoslavia* (1955). Rukavina and Hammerschmidt filed in the SAME YEAR, 1928, about the SAME TOWN, and called it two different countries. It is neither the empire at birth nor the empire at filing — it is what the man in front of the clerk said. The transcribed birth PLACE is the evidence; the birth country is not. Five of the fourteen are inside a filmed register, to the day. Otočac births are filmed 1834–1858 and all seven Otočac people were born 1888–1917, so not one is reachable. Senj births are filmed 1734–1894, which puts Mukulino (1858), Rukavina (1886), Grzin (1888) and both 1894 men — Vukelić and Hammerschmidt — in range. Marijan Vukelić is the one to open first: born 27 January 1894, the surname Milka Papić married into, and a birthplace — Klada — this archive has never had. --- Every search run is now written down — `sources/findmypast/probes-2026-09-16.psv`, 59 of them, 29 of which are zeros. A zero costs the same as a find to obtain and nothing to repeat, and the trial recorded its negatives for the same reason. The archive's other surnames are all clean negatives in the petitions. *Papich* 13 — Rudolph b.1885 was born in Italy, Stojan's oath gives no town. *Blazevich* 6 — Alice b.1894 was born in Russia. *Boras* 16 — Irene b.1884 was born in Budapest, James b.1886 in Greece. *Prpich*, *perpich*, *papic*, *vukelich*, *kalanj* all 0. The only archive surnames in this record set are Vukelić (Klada), Šojat (Senj) and Orešković (Otočac) — and all three were found by searching the place, not the name. And the village seam is exhausted. *Gospić* 1 (not our surname), *Smiljan* 0, *Karlobag* 0 — the best-read parish in this archive sent nobody — *Novi Vinodolski*, *Jurjevo*, *Selce*, *Prozor* all 0; *Ledenice* two filings by a Czech; *Povile* twenty «Povill»s; *Fiume* 516, a city and overwhelmingly Italian. Otočac and Senj are the only two that yield, and both have now been read out in full. Published: `/america` gained the third stream, the fourteen-person table with per-person reachability, and the «birth country is not evidence» section. `tools/emigrants.py` builds it from the PSV so the counts cannot drift. --- 17 September 2026 — the birthplace sweep, run through the villages the 16th did not reach. Same method, which is the only one on this site that works: the naturalisation petitions are the one record set where the birth place is INDEXED, so they can be searched by *where somebody came from* rather than by what they were called. Every non-zero below was confirmed by opening the transcript and reading «Birth place as transcribed»; nothing is taken from a results list. Nine zeros, and they are worth as much as the finds because they need never be repeated: Krivi Put, Smokvica, Krmpote, Sinac, Kompolje, Bribir, Sveti Juraj, Mrzli Dol — and that includes the two villages this family actually came from. Nobody filed a petition in the Southern District of New York from Smokvica Krmpotska or from Krivi Put. The Senj emigration went to New York; the Blažević and Prpić emigration, if there was one, did not. Three households came out, from three villages this file had never held: * KLENOVICA — Frank Pavelić, born 2 April 1894, wife Pauline b. 25 January 1903, children Frank 1921 and Pauline 1923, the Bronx. Klenovica is the Kalanj village, and the one man who filed from it is not a Kalanj. He is also entered under two surnames three years apart — *Pavelich* on the 1939 petition, PAELICK on the 1936 declaration — same birth date, same wife, same two children. That is the Žubrinić lesson again: the spelling that finds a family is the one the American clerk wrote, and an exact-match surname box will never return both halves of one man. * GRIŽANE — Kuzma Knez, born 25 September 1893, entered as *«Belgrad, Grizane, Croatia Austria»*. That is not a slip. Belgrad is the other half of the municipality's own name — Grižane-Belgrad — so the clerk wrote the hamlet and then the parish, the same shape a house number takes in a register. * GRIŽANE — Mihovil Likarić, born 29 September 1885, wife Katie, daughters Veronika 1920 and Anica 1921. One impossibility, left as written: Katie is indexed as born 25 January 1911, which would make her a mother at nine. One of the two dates is wrong, this archive does not know which, and both are copied as the index has them. Four more villages returned candidates that are NOT confirmed and are recorded as candidates: Crikvenica 4, Vrbnik 7, Brlog 4, Jablanac 2. Lukovo's single hit is fuzzy noise — the name is *Pietrzak*, which is Polish. And the URL form is written down now, because it was not, and an hour went into finding it again: the search host serves the form, `www` serves the results, and the record set is selected by `datasetname` with the set's exact printed name. A guessed slug in the path gives 404 and a guessed display name gives 500. `sources/findmypast/probes-2026-09-16.psv` --- 20 September 2026 — the set read by surname, now that the villages are exhausted. And the first thing it taught is that an index count is not a people count. `kalan` returns 20 rows and holds two households. The rows are of three kinds and only one is worth opening: | Row kind | What it gives | |---|---| | Declaration / Petition | birth date, birth place as transcribed, spouse, every child with a date | | Oath of allegiance | a name, a year and a certificate number — nothing else at all | | ids ending `/1 /2 /3 /4` | relations named *on* a record, not petitioners | Four oaths were opened before the pattern was clear. Skip them; only `/0` is the applicant, and only a declaration or petition carries a birthplace. THE KALANJ LEAD IS DEAD, and that is worth as much as a find. The trial's closing note recorded *«kalanj 0, kalan 20»* and left the twenty as a lead worth following. They are a Slovene family from Krain — Carniola — and a Polish family from Rudnyki: * Frank Kalan, b. 10 April 1873, *«Krain, Austria»*, wife Karoline b. 21 Aug 1883 * Katherine Kalan, b. 16 July 1893, *«Rudnyki, Poland»*, husband Teodor Kalanj is the largest surname in the target list — 95 of the 386 — and the one record set on this site that indexes a birthplace holds not one of them. Nobody from Klenovica or Brinje filed in these courts. That closes a line of enquiry rather than opening one, which is the whole reason the negatives are written down. And one new man. Louis Peter Vukelić, born 18 March 1892, *«Rogul, Jugo-Slavia»*, wife Mary — New Haven, Connecticut, declared 1926 and sworn 1931. He is the third Vukelić in the set and the only one not already held: Marijan of Klada-senj is in New York, and this is a different man in a different state. «Rogul» is deliberately not resolved. It is written the same way on both filings five years apart, so it is not a slip of one pen — but Ogulin and Rogulje have both been thought of here and neither is evidence. Guessing which Croatian place a clerk's spelling «must» be is exactly how this archive put a family on Krk for a week on the strength of a pencil annotation. It stands in the town column as transcribed. The sweep continued through the rest of the family's surnames, and they are other families. `papich` (13 rows, 7 petitioners, 3 opened) gives Zara — that is Zadar, in Dalmatia — and a *«Soldo or Papich»* household whose second surname is Herzegovinian. `blazevich` (6 rows, 3 petitioners) is one household, born at *«Rutkishki, Russia»*. Neither is Senj, neither is Smokvica. Four of the seven papich petitioners are unopened, and recorded as unopened rather than as absent. --- And a caution about the «birth country» column, found by accident and worth more than the searches that found it. One man gives two different birth countries five years apart: | Filing | Name | Born | Birth country | |---|---|---|---| | 1919 declaration | Riedy Papich | 26 Mar 1885 | «Zara, Austria» | | 1924 petition | Rudolph Papich | 26 Mar 1885 | «Italy» | Same birth date, same Philadelphia court, consecutive record ids. Zara is Zadar, and it passed from Austria to Italy at Rapallo in 1920 — between his two filings. So *birth country* in this index is the country that held the place when the clerk wrote, not when the person was born, and the same man is findable under two nationalities depending which paper you open. This archive already publishes a sovereignty timeline for exactly this reason — Anton Rudolf Defranceski *«was baptised under Austria, made an Italian subject at Rapallo without moving, and died under occupation»*. It now has an American index demonstrating the same thing against one man's own name, twice. --- The naturalisation set is now swept both ways, and it is exhausted. * by village — every parish and hamlet in this archive's gazetteer * by surname — `kalan`, `vukelic`, `papich`, `blazevich`, `boras`, `borash`, `sojath`, `prpich`, `papic` Everything it holds for this family is already written down. Every other hit belongs to somebody else — Slovene (Krain), Polish (Rudnyki), Dalmatian (Zara), Herzegovinian (Soldo), Russian (Rutkishki), Hungarian (Budapest), Greek (Skafers) and Ashkenazi (Sonken, Barchasch) families who happen to share a surname. And the names were not trusted to say so. *«Max Sonken or Boras»* and *«Matie Borash or Barchasch»* look Ashkenazi; *«James Boras»* looks Croatian. Judging by the look of a name is the mistake this archive keeps having to write down, so three transcripts were opened: Budapest, Skafers in Greece, and an oath with no birth data at all. The Greek one is the point — it is the record that looked most like ours and is not. Boras is a Greek surname as well as a Croatian one. What that leaves of row 57 is the 386-name target list itself, searched person by person with the given-name forms that row 60 built — which is a long sitting rather than a sweep, and the sweeps are what is finished.

Done 17 September 2026. 166 lines of hand-written prose, with the questions inlined as a `const` in the page, are now `site/src/data/questions.json` — 28 questions in the shape `kit/components/Questions.astro` reads, which had been sitting unused because only two of the seven archives held their questions as data. The shape is the point. A question is the question, what is known, what has been tried, and the record that would answer it. The middle field is the one most archives leave out and it is the one that stops a dead end being walked twice — so /open-questions/ now says, on the question about Casparus Žubrinić, that the status animarum was opened, read, and does not hold this family. That was in the prose; it was in the middle of nine hundred words. 19 open, 9 answered, filterable by state. Nineteen of the twenty-eight carry a «what has been tried». Nothing was summarised. The split was checked word by word against the prose it replaced — every word of every old body and «what would settle it», against every word of the new four fields. The first pass dropped 313 word occurrences, and two of those losses were real: the status animarum's own column list and what it was judged to cost, and the godparent column at Smokvica 114. Both were put back. What remains unmatched is 49 occurrences across 28 questions — connectives, and verbs moved into the past tense because a thing that has been tried is no longer being tried.

Done 17 September 2026. 129 lines inline, now `site/src/data/sources.json` — 12 sources, each with who holds it, what kind of record it is, what it gives, any caution, and where on the site it is used. The extraction found the page's own argument and made it a column. The page has always opened «*their weights are very different*» and then printed a state birth extract, a set of parish registers, a MyHeritage tree and a Wikipedia article as four paragraphs of similar length — which quietly says they are not. `weight` is now a field with six values, and the page is ordered by it: | | | |---|---| | state | a civil register kept by a government — 1 | | register | a parish register kept by a priest — 1 | | tree | somebody's genealogy, however carefully made — 1 | | background | history and statistics; true of a place, not of a person — 7 | | rejected | consulted, and not relied on — 1 | | absent | not consulted, and the page says so — 1 | There is no `kit/components/Sources.astro` — the row named a component that does not exist, so the page renders the rows itself. The data is in the shape one would want, so adopting it is cheap if the kit ever grows one.

Raised by a question, 17 September 2026: *«are you searching all the variations of a name?»* For surnames, yes, and it had been proved — `ZUBRINICH` returns 191 records, `ZUBRINIC` a different 119, `ZUBRINICK` ten more, one family. For given names, no. The emigration target list handed the searcher whatever form the tree held, which in these parishes is the Latin of the baptismal entry. Tested against the index the same hour: | | | |---|---| | `firstname=josip` | returns records | | `firstname=joseph` | returns records | | `firstname=josephus` | 0 | | `firstname=giuseppe` | 0 | And the site's own «include name variants» box is on by default and still returns nothing for Josephus, so whatever it expands, it does not expand across the languages a Croatian parish was written in. Josephus is the form the target list was handing people. The archive had the answer the whole time. `tools/given_names.py` has folded the forms into groups since the first week and no search has ever consulted it. What was missing was not the grouping but the language — because the question a searcher asks is not «what else was he called» but «what would the clerk in *this* index have written». --- What was built. `sources/givennames.psv` — every form, by language — and `tools/names.py`, which marks a form attested only if the records contain it and refuses to build if a form the records hold is missing from the table. 39 names, 340 forms, 159 attested. Published at /your-name/, with `?n=` deep links. The gate found three things on its first three runs, and they are the point of the row. * Two men called Mathæus were sitting outside the group of twenty-one Matijas they belong to, separated by a single «t» — `matthaeus` was in the fold table and `mathaeus` was not. The same disease as the surname fold that split on a bracket but not a slash. * The table was built from the men. Every group with a Latin feminine — *Joanna, Josepha, Georgia, Jacoba, Paulina* — had nowhere to put her, so eight Joannas, two Josephas, two Georgias and a Jacoba fell out of the fold entirely, and out of every variant list built from it. That is the worst place for a hole: the women of this tree are already its weakest part, seven direct ancestors carrying no surname at all. Five feminine groups added. And it incidentally folds `pava` to `Paulina` — the Boras entered twice that row 40 had to find without a name key at all. * The sex is counted now, not asserted. The fold declares `matija` a man's name; eight of its forty bearers here are women, one of them ahnentafel 13, Matia Pilipić, whose marriage is row 4. Croatian Matija is unisex and the table had one column for it. --- And the target list is a tool now. It said of itself that it was *«generated from the archive rather than typed, so it cannot drift from the tree»* — and nothing generated it. `tools/fmptargets.py` writes `sources/findmypast/targets.psv`: the same 386 people, plus a given forms to try column ordered vernacular and English first because the index on the other end is American or Australian, and Latin last. 305 of the 386 now carry one; 80 do not, and that is recorded as a gap in the table rather than as a claim that the name has no other forms. Latin case endings are never offered. *Gregorium* is grammar — «baptizavi Gregorium» — not a name anybody was called or indexed under, and ten rows of the old list named a person that way.

Asked for on 21 September 2026: a full Kosina narrative, to be given to the family. They had no page and could not have had one. This archive publishes eight surnames and Kosina is not among them, so all ten of them are invisible: `people.json` holds none, the family pages are generated from the surname scope, and everything known about them was scattered across /senj-1943/, the Boras material and one gravestone reading. They reach this family twice over, which is why the omission is worth correcting rather than explaining. Pavao Kosina married ANKA BLAŽEVIĆ — daughter of ahnentafel 4 and 5, Hedviga's own aunt, her father's sister. And Pavao's mother was ANGELIKA BORAS, through whom the Boras reach this archive at all. Two of the eight surnames meet in this household. And then the town was bombed. German aircraft destroyed about half of Senj on 7–8 October 1943. Pavao was killed on the second day and his stone says so in the word kept for a violent death — *pog.*, poginuo. Of a household of seven, three were dead inside thirteen months. --- 21 September 2026 — first delivery: `/kosina/` is up, built from the GEDCOM by `tools/kosina.py`. The living rule is applied by the tool and not by hand, which matters on a page about a family with survivors. Two of the ten have no death recorded and may be alive; under this estate's named-bare policy they are named and carry no date, and the tool strips the dates before they enter the data file rather than hiding them in the template. A page written by hand would have published a birth year for a woman born in 1964. What is still owed, and this row stays open until it is: Martin Kosina (b. 1841) has no father — only a mother, *Gertruda*, and no marriage, birth or death entry read for him. Otilija's death has no day, so whether she died in the raid with her father is recorded here as not claimed. Anka Kasunić, who married the elder Antun, is a name and nothing else. Oskar's wife is unnamed in the tree though their son is in it. Senj marriages are filmed to 1920 and Senj deaths stop at 1907, so the 1919 marriage is reachable and the 1943 deaths are civil-register business — row 11's letter. On outgrowing the site. The request allows for a separate site later. Nothing here forecloses that: `kosina.json` is generated and self-contained, and the page reads from it alone.

Asked for on 21 September 2026, as a second and larger piece after row 61's page. Go across and up the Kosina tree as far as it goes, find the stories, say where the family came from — and then go to the archives that hold the books and read them. Where it stands and where it stops. The tree gives four generations and stops dead at the top: Martin Martinus Kosina, born 1841, has no father — only a mother, *Gertruda*, with no dates, no surname and no marriage. Everything above 1841 is unknown, and nothing has ever been read for any Kosina: not one row of `readings.psv` names one. The first targets, in the order they are likely to pay. * Martin's baptism, about 1841, Senj. Senj baptisms are filmed from 1734 and indexed by three collections reaching back to 1692 — see /indexed/. This is a lookup, not a sitting, and it names his father. * Martin's marriage to Angelika Boras, before 1891. Senj marriages are indexed — *Hungary, Croatia-Slavonia, Marriages 1678–1927* — and a marriage entry names both sets of parents, which would give Martin's father AND confirm the Boras line in one row. * Anka Blažević × Pavao Kosina, 20 November 1919, Senj. The date is known from the family; the entry has never been read. Senj marriages are filmed to 1920, so it is inside the book, and it will give both fathers, both trades and the house numbers. * Antonius Joannes Kosina, b. 13 May 1891 Senj, d. 22 May 1925 — dead at thirty-four, married to an Anka Kasunić who is a name and nothing else. His baptism is indexed. * Otilija's death, 1943, no day. The single most wanted fact on the page: whether she died in the raid with her father. Senj deaths on film stop at 1907, so this is a civil-register request through row 11's letter, not a film read. * Neda, dead at nineteen on 7 September 1943, a month before the bombing, cause unrecorded. Same route. And where the name comes from, which is the question the request actually asks. *Kosina* is also a village in Lika, south of Perušić, and a surname along that coast. Nothing here establishes a connection either way and the archive should not assert one from a place-name — the Gologorica lesson is exactly this mistake. What would settle it is Martin's baptism, which gives a father and very often a *«ex …»* birthplace. Two archives, and one of them is already written to. Senj material sits with the Državni arhiv u Rijeci, Ispostava u Senju, and the parish books of the Rijeka archdiocese are the subject of row 52's drafted letter. A Kosina ask can ride on both rather than starting a third correspondence. What must not happen. The Kosina now publish as blood — row 63 — so two of the ten are living and carry no dates, and any research that turns up a birth year for them goes into the notes and not onto a page. --- 21 September 2026 — first sitting. The marriage is found, Gertruda is in a register, and Martin's baptism is nowhere on earth. THE MARRIAGE. «Martin KOZINA, b. Dec 1841, mother Gertruda Kozina, *no father named*» × «Angjelika Boras, of Josip Boras and Maria Kabalin» — *Hungary, Croatia-Slavonia, Marriages 1678–1927*, ark `1:1:X9DP-5Z31`. The surname is KOZINA with a z, which is why it had never been found: five spellings of *Kosina* were tried against Senj and returned nothing, because the Senj rows are filed under the Latin «Senia» and this entry carries no place at all. It surfaced on a query by given name and spouse's surname, with the surname field doing no work — the same lesson as «Piberich», a third time. Three things out of it. Gertruda is in a register rather than only in a tree. Martin has no father in the entry either — the tree's gap mirrors the document's, which points at illegitimacy or a father long dead and settles neither. And the «Dec» in both birth fields is an artefact: Martin *Dec 1841* and Angjelika *Dec 1858* is the shape of an age subtracted from a December wedding, and her own baptism says 1 September 1857. ANGELIKA'S BAPTISM, AND IT AGREES TO THE DAY. «Angelicam Boras, 1 September 1857, Senj, of Josephi Boras and Mariæ Kabalin», ark `1:1:QKMJ-SQMC`. In `readings.psv` now — her first document. MARTIN'S BAPTISM IS NOT IN THE INDEX ANYWHERE. Searched as Kosina and Kozina, births 1836–1846, country filter off — Croatia, Hungary, Austria, Germany, Bohemia, the United States. Nothing. And Gertruda appears in exactly one record on earth: that same marriage, plus an unrelated 1804 burial in Poland. TWO PATTERNS, AND THEY POINT THE SAME WAY. *Kozina* clusters in Petrovsko, Začretje, Krapina, Donja Pušća, Jalžabet, Lipovljani, Prelog — Zagorje, Varaždin, Međimurje — not Lika and not the coast. And Gertruda is a German name: with no country filter the Gertrudas of the 1840s are Prussian, German and Dutch almost without exception. This family already has two German surnames marrying in — Butvalz among the Boras, Untersinger among the Žubrinić. A third would be unremarkable, and it would also be an inference. NEXT, AND IT IS A FILM READ RATHER THAN A SEARCH. The marriage entry itself, at full resolution: a Croatian marriage register usually gives the groom's BIRTHPLACE and his father's name even where an index drops them. That one opening is the whole question. [notes/gertruda-exists-in-one-record-in-the-world.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/gertruda-exists-in-one-record-in-the-world.md)

Asked on 21 September 2026, and the question was better than it looked: «why are the Kosina outside of scope? They are still Blažević bloodline and therefore should [be in].» They were outside because scope was decided by SURNAME. `build_site_data.py` published a person when `fold(surname)` was one of the eight keys this archive carries, plus anyone on the direct ancestral line. That is how a family is FILED. It is not what a family IS — and a daughter's children take their father's name and vanish. Hedviga's aunt ANKA BLAŽEVIĆ married Pavao KOSINA. Her five children are Hedviga's first cousins. Not one of them appeared anywhere in this archive. Measured before it was changed, which is the only way to know whether an objection is a detail or a hole. Walking up from Hedviga to all 72 ancestors and down again through every descendant gives a cognatic kindred of 1,134 people. Against the surname list: > 258 blood relatives were out of scope, of whom 234 are publishable once the living rule has run — 37 Pavelići across four spellings, 11 Špalj, 9 Sekula, 7 Kosina, 6 Rivosechi, 6 Krmpotići, and twenty-two with no surname recorded at all. `ancestry.kindred()` now returns blood, and the publication test is *surname in the eight* OR *in the kindred*. Published people: 1,226 → 1,460. What it deliberately does NOT do. It returns blood only. A spouse who married in is not included unless their own surname is one of the eight — Pavao Kosina himself is still not published, though all five of his children are. That is the right boundary for a descent test, and the surname list goes on carrying the in-laws it always did. The living rule did the rest. Of the 258, 24 are living and were excluded before this change and after it. Two of them are Kosina. AND IT GREW AGAIN ON 22 SEPTEMBER 2026, so «1,460» above is history and not the current figure. Two further doors were added — `sources/documented.psv` (*a person this archive has opened a page for gets a page*) and `ancestry.parents_closure()` (*the parents of every published person, to the root, because a page that says «child of X» must be able to link X*). Published people: 1,460 → 1,903. See row 65.

Asked on 21 September 2026: «been on the MyHeritage site and I see you haven't taken them to our site?» Two were sent — Oskar Kosina's memorial card and the Otilija and Neda gravestone — and between them they answered the biggest open question on the Kosina page and corrected how this archive had been telling the story. See [notes/otilija-died-in-november-and-neda-was-the-raid.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/otilija-died-in-november-and-neda-was-the-raid.md). AND THE SAME HOUR SHOWED THAT THE PROBLEM IS NOT ONLY ON MYHERITAGE. IT IS HERE. `sources/media/manifest.psv` already listed four Kosina images, and all four were on disk in `site/public/photos/line/`: | | | |---|---| | `4505845` | Martin Kosina and Angjelika Boras Marriage — 1888 | | `4505826` | Pavao Kosina Birth — 1894 | | `4503210` | Kosina-Blazevic and Kosina-Anka Marriage | | `4505803` | Pavao and Anka Kosina Grave *(the only one any page used)* | The 1888 marriage is the one that matters, and a whole sitting had just been spent concluding it was out of reach. Row 62 ended that morning by saying the marriage entry «is the thing to read at full resolution first», and the scan had been in the library for weeks. Read, it names Martin's birthplace: «Kranjska župa sv. Martin kod Krainburg» — Carniola, the parish of St Martin near Kranj, SLOVENIA. So the job is not «download the MyHeritage photos». It is «find out what this archive already holds and has never opened». Three things to do, in order. 1. Audit the manifest against the pages. `manifest.psv` has 104 rows and the media library has 104 files. How many are cited by a page, a reading or a correction? The Kosina answer was one of four. That number, for the whole library, is the size of the problem. 2. Read the two unread Kosina scans — Pavao's birth of 1894 and whatever else the 1919 opening carries beyond entries 44 and 46. 3. Then go back to MyHeritage for what is there and not here, with the two new photographs as the first additions. A caution for step 3. Media arriving from the tree carries the living-person question with it: a photograph of somebody alive must not reach a page, and `check:living` tests the built pages rather than the manifest. Anything harvested needs the living rule applied before it is cited, not after. --- 21 September 2026, later the same day — THE THREE FILES ARE NOW IN THE ARCHIVE. They were in the Downloads folder the whole time, and this row said they were nowhere because nobody had looked there. Two came from disk at full resolution; the third was taken from the conversation itself. | mid | what | where it now is | |---|---|---| | `9000001` | Antun Kosina's portrait — cassock and Roman collar | `photos/line/`, on the Kosina page | | `9000002` | Oskar Kosina's memorial card | same, beside Argentina | | `9000003` | The OTILIJA i NEDA stone | same, in a section of its own | They are registered in a NEW file, `sources/media/local.psv`, and not in `manifest.psv`. The manifest is the output of a sweep and the next sweep rewrites it, so a hand-added row there would disappear without a word. `tools/build_media.py` now reads both. AND READING THE STONE AGAIN AT 4x CHANGED A PUBLISHED DATE. Otilija died on 11 November 1943, not the fourteenth: the first reading was taken off a phone-sized copy, and the «4» was a «1». The rule this archive wrote for itself — crop and enlarge before recording any date — was not followed the first time, on the very document that was being celebrated for settling the question. Two words of the epitaph moved with it: «svijete», a vocative, and «Raztužena», the old spelling with a z. What is left on this row is the original job, which the three files do not touch: the manifest holds 104 items and almost none of them are cited by any page, reading or correction. Audit the library against the pages, then go back to MyHeritage for what is there and not here. --- 28 September 2026 — reconciled, and it is clean. `media.json` holds 102 rows and `site/public/photos/line/` holds 107 files. Every one of the 102 rows points at a file that is published — nothing is listed and missing. The five published files with no row are the five `classification.json` marks excluded, and the reason is written there: they are stock placeholder graphics uploaded to the tree as profile pictures for people who have no photograph, one of them carrying a visible Adobe Stock watermark. Publishing those as portraits of an ancestor would be wrong twice over. So the fault this row was opened for is gone. What remains is the MyHeritage side, which is row 62's business: new uploads there still have to be noticed by a person. This row closes.

Closing row 72 opened four smaller things, and they are written down here rather than lost. 1. ~~«Kosina Dorothea — 8» is still unaccounted for.~~ CLOSED within the hour, and it cost the same correction twice. This row was written after page 8 was read at 3.2× and two Dorotheas were found — house 28 to *Primus Wit[nitz]*, house 18 to *Johann Holz*. There are three. At 8× the third reads **«6. [Hornung 1813] · 12 · DOROTHEA · ehelich · Vater: ANTON KOSINER *recte* KATZIN, *Inwohner* · Mutter: MARGARETH ZUBERTH». And it is a warning, not just an entry:** *recte* marks the legal surname against the name in daily use, so the family name is KATZIN and Kosiner is the by-name — almost certainly the house's. A Kosina in an index is not always a Kosina in a family. 2. Lukas Kosina's village is not resolved. His 1818 daughter is baptised at «Lemny» house 12 — recorded as the clerk wrote it, because it matches no village name this archive holds for the parish. His 1813 daughter's entry gives a house number and no village at all. If «Lemny» is Breg, the Marianna who died at Breg 12 in 1840 stops being *almost certainly* his daughter and becomes proved. 3. Anna 1856 and Maria 1859 vanish after 1870. Neither is in the 1836–69 death register, which is walked in full. The 1870–1890 and 1890–1911 volumes were read opening by opening for Gertraud and gave one Kozina woman, Marija at Stražišče 84 — who is of about their generation. The marriage register 1881–1899 (sig. 03919) would name both parents of both parties and has not been opened. 4. Anton 1834 and Petrus 1836 have no death. Johann and Lucia's two sons are baptised at Oberfeichting 9 and then absent from everything this archive has walked. They are the only children of that house with no end. And one method note worth keeping, now earned twice in an afternoon. Page 16 was read at 2.4× and reported empty; at 9× it gives a Kosina. Page 8 was then read at 3.2× and reported empty; at 8× it gives another. Where a surname is the whole question, the column is re-cropped before it is called empty — and a page is counted row by row, because the miss on page 8 was a third Dorothea where two had been seen. --- Later the same day — items 3 and 4 answered, and item 2 is the one still open. The two marriage registers had never been opened. Neither has an index, so both were walked: the groom's and the bride's surname columns cropped from every opening and tiled five to a sheet. sig. 03918 (1856–81) from image 58 — the year cells put that at 1874, the earliest either daughter could marry — and sig. 03919 (1881–99) in full, all 71 openings. MARIA 1859 IS PLACED, AND SHE CLOSES ANOTHER ROW OF `found.psv` WITH HER. Book p. 14: *«Nevesta: Kolodvor v Podnartu 12, rojena v Zgornjem Bitnju št. 9 · MINA KOZINA, gruntarska hči · 23»*, parents *«JANEZ KOZINA, gruntar, in URŠA SEDEJ»*. She is the «Marija Kozina — UNPLACED» this archive had carried as the only Kozina woman in twenty-five years of the 1890–1911 deaths. Two rows, one woman, and row 72's own prediction — *her marriage would settle it* — was right. AND HER DEATH IS NOW DATED WITHOUT BEING FOUND. At book p. 55: «1896, 10. februarja · Stražišče 84 · KANCIJAN ČESEN, hišar, VDOVEC» — a widower, at the house her two children died in. She died between 1893 and 10 February 1896, aged 33 to 36. And it explains the failure: that register was walked looking for *Kozina*, and a married woman is entered under her husband's name. She is in it as MARIJA ČESEN. The walk was right and the query was wrong. ANNA 1856 DID NOT MARRY IN THIS PARISH. Three Kozina brides in twenty-five years and none is her. A walked negative, not an absence of looking. AND THE SWEEP FOUND TWO KOZINA FARMING HOUSEHOLDS NOBODY KNEW ABOUT — four new people in `found.psv`: Jožef Kozina, *½ zemljak*, × Mina Kodran, whose daughter Franciška married Ignacij Kert in 1897 on his written consent; and Urban Kozina, *kmet*, whose daughter Marija married Anton Žiherl in 1898 — an entry carrying four Kozina at once, including the groom's own dead mother Marijana Kozina. None is joined to Johann's cottage by anything but the surname and the parish — except that Johann's own daughter is entered *gruntarska hči*, and he never was a *gruntar* in his life. Item 4 — Anton 1834 and Petrus 1836 — is not closed but has moved. A *«Janez Kozina, gruntarjev sin»* stands witness in 1898. Still open: item 2, Lukas Kosina's village «Lemny». --- Item 2, and it is now a better question rather than an answer. «Lemny» was described here as a village this archive could not read. It can be read. The 1799 marriage of Lukas Kosina puts him at «LEMY 37» and the 1818 baptism of his daughter at «LEMNY 12» — the same short name in two hands nineteen years apart — and the 1818 column header is «Der Orten», so it is a place. It matches no settlement in the list Matricula publishes for this parish. That list is not exhaustive: the same 1818 page carries «Prschan» for Praše, which is not in it either. So the likeliest answer is a hamlet too small for a summary list, and this has stopped being a magnification problem and become a question for somebody who knows the ground round Kranj. And one hypothesis has a number behind it. Lukas is at Lemny 12 in 1818; his daughter Marianna dies at Breg 12 in 1840. The same house number, the same family, twenty-two years apart. Either the household moved from one village's 12 to another's, which happens, or Lemy is a part of Breg under a name the clerks used and the gazetteers do not. The two words do not resemble each other, which is the weakness of it, and a house number is not a place name. --- 28 September — Marija's death is found, and the hypothesis that found it was right. The 1890–1911 register was re-walked for ČESEN after it had been walked for Kozina and given nothing. Book p. 45, entry 33: «1895, April · 2. umrla, 3. pokopana · STRAŽIŠČE 84 · MARIJA ČESEN, hišarja in sitarja žena · starost 35 let · OTROŠKA MRZLICA · previdena». She is entered under her husband's surname with no maiden name at all, and the register is inconsistent about it — *Marija roj. Eržen* and *Neža roj. Mravlje* keep theirs on the same openings — so this was not a rule anyone could have known in advance. The walk was right and the query was wrong. ***Otroška mrzlica* is childbed fever. For a married woman of thirty-five, among entries where the clerk gives grown people grown diseases, it is not a children's fever. She died in or just after childbirth, and a baptism of early 1895 at Stražišče 84 has not been looked for. And she was the fourth death in that house in three years. Kancijan 1892 at one and three quarter months; Pavla 1893 at two; ANTONIJA ČESEN, 26 February 1895, aged 1½, of *božjast* and without the sacraments — born about August 1893, the year Pavla died, and dead five weeks before her mother**. Then Marija. Her husband remarried ten months later.

Asked for on 27 September 2026, and it is right: *«you have spent so much time on Kosina but e.g. the Kalanj, Žubrinić pages are really clunky and odd — audit all the family pages considering the Kosina work and align a real story like it does.»* The audit was run against the built pages, not the source. Every figure below is counted out of `dist/`. | page | words on the page | words this archive wrote | written share | prose sections | household cards | of those, 0 children | headings that are name-soup | |---|---:|---:|---:|---:|---:|---:|---:| | `/blazevic/` | 22,915 | 592 | 2.6 % | 6 | 129 | 38 | 7 | | `/zubrinic/` | 21,733 | 663 | 3.1 % | 7 | 95 | 22 | 28 | | `/prpic/` | 12,619 | 393 | 3.1 % | 5 | 80 | 28 | 0 | | `/boras/` | 6,985 | 205 | 2.9 % | 3 | 23 | 4 | 2 | | `/kalanj/` | 23,086 | 1,191 | 5.2 % | 9 | 68 | 17 | 8 | | `/sestan/` | 6,405 | 399 | 6.2 % | 4 | 11 | 2 | 0 | | `/papic/` | 11,628 | 734 | 6.3 % | 6 | 56 | 12 | 10 | | `/vukelic/` | 2,614 | 243 | 9.3 % | 4 | 10 | 3 | 0 | | `/kosina/` | 18,488 | 18,488 | 100 % | — | 11 | 0 | 8 | So the complaint is not a matter of taste. On the eight family pages this archive's own writing is between 2.6 % and 9.3 % of what a reader scrolls past. The rest is generated listing. On `/kosina/` there is no listing that is not part of the argument. Six faults, and five of them are structural rather than a want of words. 1. Every prose section is one `<p>`. The template renders `copy.body` as a single paragraph, so a 185-word section arrives as one unbroken block. `/kosina/` uses `h3` turns, `blockquote.note` for what a register actually says, `div.note` for an aside and `div.fix` for a retraction. None of those are reachable from `familyCopy.js` at all. This is the fault that causes the feeling of *clunky*, and fixing it unlocks the rest. 2. Household cards that are not households. 126 cards across the eight have no children, and 41 have no spouse. A couple with neither is a row in «Everyone of the name», not a household, and they bury the ones that mean something — Blažević shows 129 cards, 38 of them empty. 3. Name soup in the headings. The cards print the MyHeritage display name raw: *«Mate Mathias Mathæus Kalanj Matheus Matÿ Kalanj»*, *«Joannes Juan Ivan Ive John Žubrinić»*, *«Anica / Ana / Anna "Ana" Kalanj»*. 55 headings across the eight. The tree is not to be edited — so this is a display rule, and the full form must stay on the page, quietly. 4. The cards are ordered by number of children. A page that opens at sixteen children and ends at nought has no thread. `/kosina/` moves through time, which is the only order a story has. 5. Two sections say the same thing. «Everyone of the name, generation by generation» (the descent component) is followed by «Everyone of the name» (a flat table of the same people). One of them is redundant — and `check:rows` will have an opinion about removing it, which is the point of that gate. 6. And then the writing. Only after 1–5 does it become worth asking what each page is *about*. `/kosina/` has a spine: one woman with no father in the tree, and a house in Carniola that turned out to hold four generations. Each of the eight needs its own one-sentence question and its sections put in the order that answers it. Today they are true paragraphs in no order — *«Where they are from»*, *«One hundred per cent»*, *«What is not known»*. The order of work. 1–5 are template-level and repair all eight at once. 6 is eight separate pieces of writing and gets a row each as it is started. `/vukelic/` and `/boras/` are the honest test: 243 and 205 words. If a page has little to say, it should say that in the first line rather than hide it behind six thousand words of table. --- 27 September 2026 — faults 1 to 5 are fixed for all eight pages, and fault 6 is done for the two that were named. 1–5, template-level. 44 prose sections migrated to real `<p>` tags and the block now takes `set:html`, so a family section can use what `/kosina/` uses. 55 name-soup headings became 1, through `lib/displayName.js` — a display rule only, with the full recorded form printed on the card as *«the tree records …»*; the GEDCOM is untouched. Cards now run oldest first (`sort="given"`, which the kit already had). The childless and sole-parent cards are kept and counted, with a note saying they are a fact about the tree and not about the family. The second «Everyone of the name» is now «Every person of the name, with where they were born». 6, the writing — two of eight. | page | story words before | after | structure now | |---|---:|---:|---| | `/kalanj/` | 1,191 | 1,936 | 9 sections · 26 paragraphs · 1 blockquote · 3 notes · 1 retraction block · 2 sub-turns | | `/zubrinic/` | 663 | 1,425 | 7 sections · 15 paragraphs · 1 blockquote · 4 notes | What changed was mostly the order. `/kalanj/` opened on house numbers and buried its discovery — *they are not a Klenovica family, they are Brinje people* — in seventh place. It now opens there, with the Latin quoted, and the coast material follows as *what the family looked like after they arrived*. `/zubrinic/` opened on «Where they are from» and kept its real subject — 422 records, 90 loose ends, and three Frontier offices the tree has never carried — for last. Both now end with «What is not known, and what would answer it», which is the section `/kosina/` has and none of the eight had. Still to write: `/blazevic/`, `/papic/`, `/prpic/`, `/boras/`, `/vukelic/`, `/sestan/`. `/boras/` (205 words) and `/vukelic/` (243) are the honest test: a page with little to say should say so in its first line. --- Later the same day — fault 6 is done for all eight. Every page now opens on what it is about, moves in an order, and ends with «What is not known, and what would answer it» — the section `/kosina/` has and none of the eight had. | page | story words before | after | structure now | |---|---:|---:|---| | `/kalanj/` | 1,191 | 1,936 | 26 ¶ · 1 quote · 3 notes · 1 retraction · 2 turns | | `/zubrinic/` | 663 | 1,425 | 15 ¶ · 1 quote · 4 notes | | `/papic/` | 734 | 1,326 | 19 ¶ · 1 note | | `/blazevic/` | 592 | 1,219 | 14 ¶ · 1 quote · 2 notes · 1 retraction · 1 turn | | `/prpic/` | 393 | 1,044 | 12 ¶ · 1 quote · 2 notes | | `/boras/` | 205 | 1,002 | 9 ¶ · 2 notes | | `/sestan/` | 399 | 973 | 11 ¶ · 1 quote · 1 note | | `/vukelic/` | 243 | 856 | 10 ¶ · 1 note | | | 4,420 | 9,781 | | What actually changed was the order, and then the honesty of the ending. `/kalanj/` buried *they are not a Klenovica family* in seventh place. `/blazevic/` kept *the parish is named and nobody has opened it* as a mid-page aside; it is now the first thing in what is not known, because five of the eight generations on the line this archive is named for sit in that one unread book. `/prpic/` had Antonija's burial at Mirogoj as a closing shrug; it is now a question with two others beside it. `/sestan/` already had the best spine of the eight and needed only the label quoted where a reader can see it. And `/boras/` gained a section this week's other work earned. Angelika Boras married Martin Kosina — which was a line in a household list until the Carniolan household was read this month. Two infant deaths in 1896 at Senj, «Martin Kosina × Angjelina Kosina», that the family tree does not have. One typed number was introduced and then taken out again. A rewrite of `/papic/` said *seventy-five people are written Papić (Papa)* — true on the day and exactly the fault row 68 closed. It now points at `/name/`, which counts them on every build, and says so. --- 28 September — the `ARCHIVE_OUT` audit this row's near-miss demanded, finished. Every tool in `tools/` was checked for a hard-coded `site/dist`. Three had one, not two: `searchindex.py` and `check_site.py`, both fixed on the 27th, and `tools/sitemap.py`, which turns out to be dead code — the build has called the kit's `sitemap.py` for weeks, and that one resolves through `kit/tools/outdir.py` and honours `ARCHIVE_OUT` correctly. The local copy is not deleted (this estate keeps what it replaced and says why); it now refuses to run and points at the kit's. And the kit had already solved this and written it down. `kit/tools/outdir.py` says the fault «reached six of the eight» archives and that `ARCHIVE_OUT` deliberately beats `--dist`. So the `--dist "${ARCHIVE_OUT:-dist}"` added here on the 27th was belt and braces rather than the fix — but it is the form the kit names approvingly, so it stays. A new gate, `check:datalinks`, closes both link faults that reached the published site. One: a raw HTML anchor with a root-relative target, written into a work-list note. The kit's `WorkList.astro` renders notes with `md()`, whose single link rule rewrites a Markdown link through `u()` and passes a raw anchor through untouched — so the anchor ships without the base path and 404s. That is exactly what failed the deploy of commit `4c72556`. Two: a link to a `/found/<slug>/` that no longer exists, which is what a renamed row does silently. The gate was written wrong first and that is worth recording: it flagged 65 raw anchors across `corrections.json`, `questions.json` and `sources.json` — all of which are rendered by components that apply the base themselves, so all 65 work. A gate that reports 65 working links to catch one broken one gets switched off. It is now scoped to `worklist.json` alone, with the reason in the file, and it was tested in both directions: it passes clean and it fires on an injected anchor. It runs in under a second against the sources and needs no `dist` at all.

The account holder was given a Trove API key on 27 September 2026, valid to December 2026. It is the first keyed API this archive has had. The key is not in this repository and must never be. It lives in `sources/keys/`, excluded entire by `.gitignore`. This repository is public and publishes to GitHub Pages: a key in a tracked file is a key on the open web inside two minutes. `tools/trove.py` reads it from there and refuses to run without it rather than falling back to anything. What it is worth. One branch of this family has been read only through indexes: Ivan Žubrinić reached Port Pirie about 1884 and his descendants are registered Zubrinich through Broken Hill, Whyalla and Quorn to this day. Trove is the newspapers those people lived in. The first sweep produced ten verified articles and they are on /port-pirie/ — a Land Board block in 1893, a road described by his name in council minutes in 1897, a prosecution under the Education Act in 1900 in which a daughter tells the court her sibling has had *sandy blight* for eighteen months, a greyhound called Kaiser coursing at Adelaide in 1901, and a pioneers' reunion in 1911. The method, because it is not obvious. The API enumerates and returns no article text with this key — `include=articletext` and `reclevel=full` were both tried and both give an empty field. The OCR comes from the free public page instead. And the search is fuzzy: 9,263 hits for *Zubrinich*, of which the earliest are advertising columns containing no such name. 36 of the first 40 South Australian hits are real. The total is never the finding. What to run next, in order. 1. The rest of South Australia — the sweep stopped at 40 of 8,790 and everything after 1911 is unread. 2. Broken Hill and Whyalla, which are New South Wales and South Australia respectively and were not in this query at all. 3. Death and marriage notices for the whole family, which are what fix dates. 4. The other spellings — *Zubrinic* (261), *Xubrinich* (13), *Zubrinick* (52) — none yet run. 5. And the other eight surnames. Trove is not only Žubrinić: Kalanj emigrants, the Senj Kosina who went to Argentina, and any Blažević who reached Australia are all searchable now and none has been looked for. --- 28 September — items 2, 4 and 5 of the five done, and the table is now data. Broken Hill (item 2). New South Wales had never been queried. 23 of the first 24 verified, and it is the *Barrier Miner*: M. Zubrinich and «Weedie» Zubrinich for Solomontown in 1922, and L. J. Zubrinich running the football and cricket socials at Waverley Hall as honorary secretary. A nickname is something no register ever gives. The other spellings (item 4) are not spellings. *Xubrinich* and *Zubrinick* were run in full and are all Port Pirie or Crystal Brook, 1922–1938 — and one 1925 paragraph carries both «Mrs. M. ZubrinIck» and «Mrs. Zubrlnich». The scanner, not the family. Important, because X is the Venetian convention in 1834 and OCR error in 1925, and this page was one sweep from counting them together. The other surnames (item 5) were probed and left. Kosina 1,923 · Prpić 991 · Papić 676 · Blažević 418 · Šestan 244 · Kalanj 123 — all fuzzy totals, none with a known Australian branch to anchor a sweep. Žubrinić got the sweep because Žubrinić has a family there to find. And the hand-typed table is gone. `sources/trove.psv` + `tools/trovedata.py` → `trove.json`, wired into `regen.py`; the page says «{trove.count} articles» by counting rows. That was the fault row 68 had just closed, reintroduced within a day, and caught. Still to run: the rest of South Australia (the sweep has read 40 of 8,790), Whyalla and Quorn, and death and marriage notices, which are what fix dates and are the reason to keep the key. --- 28 September, later — Family Notices run, and it gave what the social columns couldn't: spouses. 231 hits in this category alone, 50 of 50 fetched verified — the cleanest hit rate this row has seen. Three men this tree has carried for years with no spouse recorded — Michael Valentine, John Stephen and Leonard Francis Zubrinić — turn out to have married, and none of their wives (Minnie Doris, Irene Dorothy Patten, Asta Louise Hudson) nor an infant son (Leonard John, d. 1919 aged 5 weeks) were anywhere in `people.json`. Four new `found.psv` rows. And it corroborates the tree to the day on two deaths already dated, adding the manner (Kate Zubrinić: *«accidentally killed»*, 1927) that no tree field carries. Still to run: the other two Family Notices hits beyond the first 50, and the same category for New South Wales (Broken Hill), which has never been queried at this category level — only the general search. --- 28 September, later still — New South Wales Family Notices, first time run. 28 of 29 verified. Gave Leonard James Zubrinić (1894–1985, WWI veteran, tree had birth and death and nothing between) his whole Broken Hill family: wife Eva Uhe, three married daughters, a grandson. Four more found.psv rows. And a genuine open lead rather than a closed one: «Kate and Jack Zubrinich» of an 1917 Broken Hill notice may be John William Zubrinich × Catherine Lucy Murphy, already in the tree — Kate/Catherine and Jack/John both plausible, the surname exact, nothing else yet checked. Worth a deliberate sweep of its own: the Barrier Miner/Barrier Daily Truth Murphy-family notices around 1914–1921, and whichever record gave the tree "Catherine Lucy Murphy" in the first place, to see if it already carries a birthplace or parents that would settle it. What remains untried in this row: the other three spellings (Xubrinich, Zubrinick) at the Family Notices level specifically — only run against the general search so far; Victoria and Western Australia, queried once each for a bare total and never opened; and every other surname this archive holds, none of which has been searched on Trove at all.

Read rather than assumed, on 28 September 2026. Johann Kozina's second wife is the last unknown woman in the Carniolan household: she bears Franziska 1853, Anna 1856 and Maria 1859 at Oberfeichting 9, and no record yet read names her parents. The window is tight and it was walked. Lucia Labornik died 30 November 1850; Franziska was baptised 19 September 1853 and entered *ehelich*. The Kranj–Šmartin marriage register 1816–1855 was read across book pages 66 to 71 — April 1850 to February 1853, without a gap, about forty marriages. There is no Kozina on any of them. So the finding is about her. A marriage is celebrated in the bride's parish. Johann's first marriage is in this book because Lucia Labornik was local — Kranj–Šmartin, January 1833. Ursula Sedej was not. That is a fact about her origin obtained entirely from where her marriage is *not*. What to do. The neighbouring parishes are on Matricula, free, no login: Stara Loka, Kranj, Šenčur, Preddvor, Naklo. A Sedej marrying a widowed cottager of Zgornje Bitnje between 1851 and 1853 is a short walk in each. Stara Loka first — the 1890s deaths at Kranj–Šmartin repeatedly note people *«doma iz fare Stara Loka»*, so that is the parish this one exchanges people with. And one practical warning. Matricula was slow the night this was read: several image fetches timed out and needed three attempts. Not a block, but budget for it. --- Answered the same day it was opened, and from the other end. This row said her marriage would be in her own parish's book and named Stara Loka as the place to start. It did not need that book at all. The Kranj–Šmartin marriage register 1856–1881 was walked in full for a different question — Anton 1834 and Petrus 1836 — and five openings from the end, at book page 54: > 28 May 1873 · Ženin: Zgornje Bitno 72, MARTIN BRENK, *gost*, 48, vdovec > Nevesta: Zgornje Bitno 9, URŠULA SEDEJ, po ranjcim možu KOZINA, POSESTNICA, 46, vdova > *Roditelja nevestina:* ANDREJ SEDEJ, GRUNTAR NA NOVI OSLICI, in JERA ROJE Johann's widow remarrying. And a second marriage names parents exactly as a first one does. Her age — 46 — puts her birth at about 1826–27. Her parents are Andrej Sedej and Jera Roje. Her origin is NOVA OSELICA, in the Poljane valley south-west of Škofja Loka, twenty-five kilometres off and in another parish — which is why her first marriage was not in this book, exactly as this row predicted. **And she is *posestnica* of house 9**: Johann died a *Häubler* in 1868 and five years later his widow holds the property in her own name. That last fact settles the contradiction row 73 flagged and could not explain — *gruntarska hči* at Mina's wedding in 1883, against a father who was a cottager all his life. Her mother's father was a gruntar and her mother was a posestnica. What is open now is one parish further out. Nova Oselica's registers hold Ursula's baptism of about 1826–27 and the marriage of Andrej Sedej and Jera Roje. Not one page of them has been opened.

Asked on 28 September 2026: *«is everything in our site? e.g. [notes/the-kozina-of-zgornje-bitnje.md](https://github.com/daviddef/TheBlazevic/blob/main/notes/the-kozina-of-zgornje-bitnje.md) — is there a reason we're creating more docs or should it be in our site?»* The answer was no, and the question was right. How notes work here, which is not obvious. There is no `/notes/` route. A note is a working paper in the repository and a reader meets one only when a page cites it — `notelinks.py` turns a bare `notes/` + filename in a page into a GitHub link. The site carries the argument; the notes carry the working. That is a deliberate design with one failure mode, and this archive was in it. 76 notes on disk. 59 cited. 17 reachable from nowhere — among them a 4,769-word guide to the Croatian state archive's register search, a 2,559-word account of the evening that produced the Kozina household, and the whole of the 1888 Senj marriages read end to end. Fixed, and the fix was mostly not new writing. Six of the ten substantive orphans already had a row in the search register that simply did not name them — Senj 1888, Novi Vinodolski 1650, Grižane 1738, the FamilySearch catalogue, the film-structure lesson, the harvest map. ARHiNET had no row at all, so the whole source was in a note and nowhere else; it has one now. `uroda-1791.md` became a row in `sources/readings.psv` — *«Dominus Georgius Uroda, 10 September 1791, buried in his own crypt»*. The namefold note is cited from `/name/` and the K-section note from `/kosina/`. And the categories are now declared rather than assumed. `notes/README.md` sets out three kinds — notes cited from a page, `PROMPT-*` briefs written for another archive's session, and hand-offs whose contents must not be published here (`for-the-defranceski-archive.md` and `the-kozina-of-zgornje-bitnje.md` exist precisely to keep those lists off this site). `tools/notescheck.py` runs in the build and fails when a note is reachable from nowhere and has not been excused with a reason. And the same audit caught two findings of this session that had not reached a page either — that Franz Kosinar is absent from the 1603–1779 baptismal index, and that a Dolenc stands at four baptisms of this household across forty-one years. Both are on `/kosina/` now. Checking was the point; twenty findings were verified against the built pages and eighteen were already there.

Opened 28 September 2026, the same evening it was proposed. Ursula Sedej's 1873 marriage names her father as a *gruntar na novi Oslici* — Nova Oselica, in the Poljane valley. Eleven volumes, confirmed free on Matricula, no login. Neither register has a name index, so this was a straight walk rather than a lookup, and it did not reach the end. Confirmed: Sedej is a real, resident family at Nova Oselica. Gertraud Sedej and Catharina Sedej are mothers of two different brides marrying in January 1827; Agnes Sedej stands godmother the same years. And an Andre Sedej appears as a father — *½ Hüblor*, wife Maria Briz, a child baptised about 1829–30 — but the wife's name does not match Jera Roje, so this is either a different man of the same name and parish, or a first wife who died before Jera Roje married him. Not settled either way. What was walked, so nobody repeats it: baptisms sig. 01584 images 3–25 (1826–c.1832); marriages sig. 01589 images 38–44 (its final pages, June 1824–January 1826); marriages sig. 01590 images 1–8 (1826–1828). No Ursula Sedej baptism and no Andrej Sedej × Jera Roje marriage in any of it. What is still open, in the order worth trying: 1. ~~Sig. 01584 past image 25~~ EXTENDED to image 36 (roughly 1830–1834), same session. Six more Sedej women turn up as mothers (Mariana, Gertraud twice, Maria) and three more Sedej men as godfathers — the family is thoroughly resident, four women and three men now attested. Still no Andrej Sedej as father and no Jera Roje as mother. And Roje is now independently confirmed too — image 36 names a mother «Agnes Roj[s]ka», the adjectival form for a woman surnamed Roje; a different, unrelated woman, but it means both halves of «Jera Roje» belong to this parish. Sig. 01584 past image 36 is what remains of this item. 2. Sig. 01589 before image 38 — the bulk of the 1785–1826 marriage volume, if Andrej and Jera married well before her birth. 3. Sig. 01590 past image 8 — if their marriage falls just after 1826. 4. Sig. 01583 (baptisms 1785–1826) and 01588 (marriages 1763–1784) — the two volumes before those already opened. A method note, since this parish has no index. With no name index and no way to jump to a year, the fastest read is the tabular baptismal grid rather than the Latin prose marriage register — the marriage book's early volume (01589) is running prose, one sentence per couple, and far slower to scan than a column.

A cross-session message from the estate's landing session on 28 September 2026 flagged a fault class: a finding recorded in a log can sit on disk while a person's page still says nothing is known. It had already reached five archives once, at Mazza, each porting the lesson into its own `check:evidence`-style gate rather than a shared file — the right pattern, since the message said so itself and named the reason: a shared checker would get each archive's data shapes wrong. This archive and D'Arcy were the two with no such gate. It was not hypothetical here. An hour earlier, checking rather than asserting an answer to *«are all your findings updated?»* had found exactly this: Ursula Sedej was fully written into `/kosina/`'s prose across four commits and never added to `tools/kosina.py`'s hand-typed household table — the data every count and stat line on that page is built from. The page said *«6 read out of the parish registers»* directly above prose describing a 7th. Investigated before building anything, per the peer's own caution against guessing at another archive's data shapes. The rest of the pipeline turned out to be structurally immune: `readingsdata.py` builds `readings.bySlug` from every row of `sources/readings.psv`, consulted directly by every `/people/<slug>/` page; `/found/<slug>/` is generated one page per `found.json` row, so a row cannot exist without a page; and the eight core family pages call `descentOf()` in `lib/households.js`, which pulls in every `found.psv` row by surname — a computed join, not a maintained list. Verified both empirically: every reading citation and every found citation was checked against the actual built HTML, not just the data files. Clean, zero gaps. Kosina is the one exception, and it says why in its own code comment: it isn't one of the eight surnames `descentOf()` joins on, so `tools/kosina.py`'s CARNIOLAN and HOUSEHOLD tables are hand-typed, deliberately — *"a claim about people … should be legible as one, not inferred from a string that might drift."* A hand-typed list has no mechanism to notice when `found.psv` grows a row that belongs in it. That is exactly where the fault landed. `tools/householdcheck.py`, tested both ways before it shipped. Reverting Ursula Sedej from `kosina.py` and running the gate reproduced the exact fault and failed on it; restoring her passed clean. The match rule is narrow on purpose, after the `datalinks.py` lesson about pattern-matching that flagged 65 things and broke none: a given name alone is too common to mean anything, so a Kos-/Koz- surname stem must also appear in the same relation sentence — and the stem, not the exact spelling, because Kosina/Kozina/Kosinar/Kosiner are the same clerks' hands on the same family, which is the subject of the page's own section on the spelling. One legitimate exception found and excused, not fixed, because it isn't a fault: `Marija "Mina" Kozina` and `Maria Kozina of 1859` are two `found.psv` rows for one woman by design — the page tells the story of two lines of evidence, the marriage register and the baptismal index, converging on one identity. Both slugs are linked from the prose. Excused by name in `EXCLUDED`, with the reason on the line, the way `notescheck.py` already does it. And the finer distinction the peer's message named — unmeasured anywhere yet, by its own account — is what this gate actually tests. Not *does a person have a page* (every found.psv row already does, by construction). *Does a person the household's own words name actually appear in the household* — the flagship narrative page, not a page a reader would only reach by searching. Wired into `npm run build` as `check:household`, last in the chain.