Reading the archive against itself
Six open questions, answered or half-answered out of material this site had already published. No subscription, no archive, no letters — just reading what was written down.
Yesterday this archive discovered it had been calling Constantine D'Arcy unplaced on three pages while publishing his christening on a fourth. The obvious next question was how many of the other open questions were in the same condition. The answer is most of them.
A tool, not a hunch
tools/crossread.py holds the open questions as search terms, reads every data file and every built page of this site, and prints each place the terms appear — excluding the page that owns the question. It finds nothing new in the world. It finds what was already here and not being read.
719 data files and pages. It runs in under a second, and it should have existed in July. tools/crossread.py
Robert and Jean had ten children
This site's page on Robert D'Arcy's family listed FIVE, and added a sixth yesterday. The family tree it is built on holds TEN, and the Hornby page has said “ten children” in plain text for weeks. Three of them were born in Barbados and three of them died in Sussex.
| Child | Life | Born | Died |
|---|---|---|---|
| Joseph D'Arcy documented | 1780–1848 | 14 Mar 1780 Portsmouth | Lymington, Hampshire |
| Margaret Maria Isabella D'Arcy documented | 1781–1843 | 26 Sep 1781 Portsmouth | Le Havre, France |
| George Pitt D'Arcy documented | 1783–1849 | 18 Feb 1783 Portsmouth — “or West Indies” | Parramatta, New South Wales |
| Richard d'Arcy inferred | b. 1784 | 1784 Portsmouth | — |
| Constantine D'Arcy documented | 1786–1805 | Feb 1786 BARBADOS | Chatham, Kent |
| Jane D'Arcy tree | 1787–1875 | 15 Dec 1787 BARBADOS | FALMER, SUSSEX — aged 88 |
| Catherine D'Arcy tree | 1789–1876 | 10 Dec 1789 BARBADOS | BRIGHTON — aged 86 |
| Anna Maria D'Arcy tree | 1791–1791 | 16 May 1791 — | died 24 October, aged five months |
| Anna Maria d'Arcy tree | 1792–1800 | 10 Oct 1792 Portsmouth | died aged seven |
| Charlotte D'Arcy tree | 1794–1869 | 19 Oct 1794 LYME REGIS, DORSET | BRIGHTON — aged 74 |
And the West Indies posting is bracketed without a single new record
The Corps page once showed what it called «a blank of twenty-four years» in Robert D'Arcy's career, between his commission of 17 January 1776 and his majority of 1802. The birthplaces of his own children close most of it.
| 1783 | George Pitt | Portsmouth — “or West Indies”, says the tree, hedging |
| 1784 | Richard | Portsmouth |
| Feb 1786 | Constantine | BARBADOS |
| Dec 1787 | Jane | BARBADOS |
| Dec 1789 | Catherine | BARBADOS |
| Oct 1792 | Anna Maria | PORTSMOUTH — he is home |
| Oct 1794 | Charlotte | Lyme Regis, Dorset |
Robert D'Arcy has a brother nobody here has ever mentioned
The Hornby page argues at length about Robert's parentage — the Earl of Holderness, the inserted “Thomas (Francis) D'Arcy”, the tree contradicting itself. It has never once said that the same tree gives Robert a SIBLING, hanging off the same two parents.
The tree's note on him reads: “Lt in the Royal Artillery. died s.p. 01/06/1778 appt to De Lancey's Brigade (Cornelius Dersey! original transcription). Probably about 19 yrs old.” De Lancey's Brigade was a Loyalist corps raised in New York for the American war. And the tree's note on ROBERT says he was “wrecked on his voyage to America to fight as a volunteer in the American war”, losing the papers that named his parents.
Lydia has a birthplace, a death date and two parents
This site says of Hannah Saniger's mother: “FamilySearch's shared tree gives John Saniger a wife LYDIA JONES, 1770–1851 … It is not sourced.” True as far as it goes. What the page does not say is how much detail sits behind the name in this archive's own data.
| Born | 13 March 1770, WOTTON-UNDER-EDGE, Gloucestershire |
| Parents | WILLIAM JONES and SARAH |
| Married | John Saniger, 1797, Berkeley |
| Died | 7 JULY 1851, BERKELEY, Gloucestershire |
So does Catherine Cotton
Listed as an open question — “where did Catherine Cotton come from?” The tree answers it in four fields: born 11 OCTOBER 1746 AT BERKELEY, died 27 SEPTEMBER 1778 AT BERKELEY, married THOMAS SANIGER at Berkeley on 10 APRIL 1769, parents EDWARD COTTON and ELIZABETH. Berkeley St Mary's transcript stops in 1677, which is why nobody found her — but a marriage of 1769 and a burial of 1778 are within reach of the Gloucestershire Archives' own holdings, and the question was never framed that way.
And Martha Wakefield, which went the other way
This site says Martha “may never have existed”: named on five pages, searched at Bristol under every spelling and every record type, and found nowhere. The crossread points at two things the Bristol search could not have reached.
| The tree says she died in QUEENSLAND, not England | 23 April 1899, Maryborough — buried Maryborough Cemetery, Plot Monumental L, Grave 410 |
| And it says she married | Its source is cited as “Backhouse Web Site — MARTHA HURFORD (born Wakefield)” |
So the search was run in the wrong country, under the wrong surname, for a woman whose own tree entry gives a death date, a cemetery and a plot number. That took ten minutes to test in the free Queensland index, and the answer is better than either side of the argument.
The tree's date, to the day. And the index names her parents: FATHER JOHN WAKEFIELD, MOTHER a HUGHES.
The other candidate — a Martha Wakefield did marry Philip Millingen in Queensland in 1887. Her parents are HENRY WAKEFIELD and CATHERINE DUNCAN.
And no Wakefield death anywhere in the index gives the parents James Wakefield and Hannah Saniger for a Martha.
And the lesson had already been filed once
On 13 September this archive published an error about Jabez and Zillah Wakefield, announced as “two children nobody here had ever heard of” when both were on a passenger manifest it had transcribed itself in July. The lesson recorded at the time reads: “Check a discovery against what the archive already holds before calling it new. The tree is not the archive.”
Three of these are no longer open
This page was written on 13 September 2026 to say that six questions filed as open were answerable from what the archive already held. On 14 September three of them were carried further — not from the archive's own files this time, but from free sources nobody here had opened.
LYDIA SANIGAR, 1851, register 3, entry 265, in the free Gloucestershire Registration Service Index. The date the tree gives — 7 July 1851 at Berkeley — sits inside a district whose deaths are complete from 1837. She was invisible under SANIGER and present under SANIGAR. The certificate is now orderable against a local reference; her maiden name still needs it.
PROB 11/1056/294 — the will of BENJAMIN COTTON, cordwainer of Berkeley, proved 16 September 1779, eleven months after Catherine was buried in that parish. Her father is given as Edward, so Benjamin is not him; but this archive had never held a single record from the Cotton side, and a Berkeley Cotton's will is where her children would be named.
This page bracketed the West Indies years from the birthplaces of Robert D'Arcy's children, with no record of the posting itself. The Corps' own history carries “ROBERT D'ARCY, BARBADOS” in a station list — and puts him at the siege of Fort St Philip's on Minorca in 1781, inside the blank this archive called empty.
And a second tool, for when the archive disagrees with itself
This page came out of a filing failure: a fact published on one page while three others called it unknown. A week later the same shape appeared in a different form — a DATE printed as 1776 on five pages and 1778 on a sixth, for weeks, with nothing in the build able to notice.
So there is now a second script. crossread.py reads this archive against its open questions; agree.py reads it against itself. For every person it names often and every event word that carries a date, it collects every year appearing near both and reports the pairs carrying more than one.
What it found, and what it did not
Then thirty-one false alarms, all triaged
Every remaining pair was checked by hand and every one is a false positive, in three kinds. REPEATED FORENAMES: this family has a Thomas Saniger in 1677, 1744, 1769, 1803 and 1851, and a name-matcher sees one man with five dates. ADJACENT EVENTS: a marriage of 1779 sits one line above a baptism of 1780, and proximity cannot tell a neighbour from a contradiction. A BIRTH AND A LATE BAPTISM: “29 Apr 1825 / bapt. 31 Mar 1826” is one child, correctly recorded, eleven months apart.
And one thing it had to be taught
This archive states wrong dates ON PURPOSE. Its method is to print what was rejected — “the tree says 10 March 1824, ten months early” — so a page CORRECTING a date reads exactly like a page asserting it. The first version duly flagged /jean-ward for 1777 in a sentence that existed to rule 1777 out. Windows carrying a negation, or the archive's own correction idiom, are now skipped. the method defeats the checker, until the checker knows the method
What this cost
Six of the questions this archive has been carrying were answerable, or half-answerable, out of material it had already published. One of them it had published on the same page as the question. None of this required a subscription, an archive, or a reply to a letter. It required reading what was already written down, which is the one research method that costs nothing and the one this archive had not applied to itself.
The errors this produced → · What is still open → · What it would take to settle them →