What changed

The whole site read against itself, and eleven findings turned out never to have reached a page

did

• DAVID ASKED FOR THE WHOLE SITE TO BE READ — people pages, visuals, trees, writeups, snapshots and links — AND THE GATES ALL SAID OK BEFORE ANY OF THIS WAS FOUND. That is the useful part. Every gate passed on the first run of the audit, and eleven findings were still sitting in the data without ever reaching the page they belong to. A green build means no rule was broken; it does not mean the archive says what it knows.
• THE FIRST THING CHECKED WAS THE WRONG THING, AND IT COST AN HOUR. Four relationships documented this week looked absent from the charts, so they were added to the edge list and the result was verified against `register.json`'s `linked` — which is a COUNT OF ENTRIES CARRYING A LINK and has nothing whatever to do with relationships. The edges live in `provenance.json`. The tool had reported its own failure plainly, in its own output, and it was not read: NINE REL NAMES THAT DREW NO EDGE, printed as a bare list for weeks.
• AND THE REFUSALS WERE RIGHT. Of the nine, two were ambiguous names — two John Atwells, three Richard Atkinsons — and the tree settled both by its own id links: James Atwell's parents are John Atwell and Maria Rossiter, and John Atwell born 1824 lists that same James by identifier as his child. Those two now draw, with the Queensland death registration as the reason. The other seven cannot ever draw, because the person at the far end exists only in a document and a chart draws the tree. FOUR OF THEM WERE THIS SESSION'S OWN, ADDED ON A WRONG DIAGNOSIS, AND THEY ARE GONE AGAIN.
• A LIST OF NAMES IS NOT A DIAGNOSIS, so the reporter now says WHY each name fails — ambiguous, document-only, or absent — because the three want three different answers and a bare list invites none of them.
• ELEVEN ROWS CLAIMED A PAGE THAT HAD NEVER CARRIED THEM. The `published` gate counts these and had been printing «11 quiet» without anyone acting on it. They were not noise. The Moreton Bay Courier had been read for the General Hewitt and had NOT confirmed the arrival — an honest negative result that no reader could see. The London Gazette had been checked for Vivian Sneyd's first Military Medal and gazettes it without a citation. The National Library of Wales serves a bot check over a digitised will, and `/behind-the-paywall` — the page about walls — did not mention the one wall that is not money. FreeREG offers County Limerick with no data behind it, and `/atkinson` did not say so. All eleven now reach their page, and the count is 0.
• TWO OF THEM WERE FILED ON A PAGE THAT RENDERS A DIFFERENT FILE. `/coverage` renders `indexes.json`; `coverage.json` renders on `/searched`. Two findings about Gloucestershire record classes pointed at `/coverage` and could never have appeared there whatever was written. They are re-filed on `/swonhungre`, where the surname they are about lives.
• AND ONE WARNING WAS THE INSTRUMENT, NOT THE ARCHIVE. The errand for the will of THOMAS SANNIGER, Able Seaman of HMS Eurydice, was graded on the single word «Seamen» — taken from «Seamen's wills are short» — and reported as unpublished on a page that names SANNIGER three times. The test asks the finding first, which is right, and then refused to look at the identification beside it, which is not. ONE COMMON WORD IS NOT A DISTINCTIVE TOKEN: below two tokens it now reads both fields. Same fault as the version that certified a row because a page said «Wales», pointing the other way.
• A FIGURE IN PROSE GOES STALE THE NEXT TIME THE TOOL RUNS. The snapshot count was written into a sentence as «70 of 84» and the instrument now says 73 of 87 — one week. `snapshot.py` writes the number where the page can read it, and `/sources` carries it with the hosts that decline to be archived named, so the page cannot disagree with the tool. The dated changelog entries are left as they were: they were true on the day and rewriting them would falsify the record.
• WHAT THE AUDIT DID NOT FIND IS WORTH SAYING. No page is without an inbound link; no asset reference is broken; no citation this archive chose for itself is dead, and the nine dead ones are all somebody else's footnotes inherited from the GEDCOM export. 858 pages built, every gate green, 248 record lines reaching 62 of 62 people in the tree.

found Fixedtwo recorded relationships that had never drawn: John Atwell and Richard Atkinson were each AMBIGUOUS in the tree, and the edges now carry the year that settles themCheckedfour relationships added this week were REFUSED by the chart and the refusal was right — the second party in each exists only in a document, and a record line is not an edgeAddedthe Moreton Bay Courier of 23 December 1854 was read for the General Hewitt and DOES NOT CARRY HER — the arrival date rests on the register alone, and now says soAddedthe first Military Medal has no citation anywhere: the Gazette promulgates the award and never says what it was forAddeda wall that is not money — the will of a Chepstow mariner, digitised at the National Library of Wales behind a bot check this archive will not passAddedFreeREG offers County Limerick and holds NOTHING behind it, so every free Limerick search was searching a county that is not in the indexAddedtwo record classes that cannot be searched by name at all — feet of fines catalogued at piece level, and the Gloucestershire editions in copyrightAddedthe coverage verdicts and the Trove gazette key, both of them faults in this archive's own instruments, now published on the page whose numbers they affectedFixedthe snapshot coverage figure is computed by the tool and read by the page, after a sentence saying 70 of 84 survived a week into the instrument saying 73 of 87Checked0 pages with no inbound link, 0 broken asset references, and 0 of this archive's own citations dead — the nine candidates are all inherited from the GEDCOM export

The Sanigars, Sneyds and Wakefields get the same treatment, and Oswestry hands over a family

did

• DAVID ASKED FOR THE OTHER THREE LINES AND OSWESTRY ANSWERED FIRST. The Shropshire Parish Registers for the DIOCESE OF ST ASAPH are printed, free and out of copyright, and eight volumes were downloaded whole and read. They give DAVID OLIVER, COOPER, OF CROSS STREET, OSWESTRY — an occupation and an address this archive did not have — BURIED AT OSWESTRY ON 21 JUNE 1808, AGED 59, which puts his birth about 1749. With Mary he baptises DIANNA on 23 MARCH 1777, ELIZABETH on 19 MAY 1782 and DAVID on 12 JULY 1789. Elizabeth is this archive's Elizabeth Margaret Oliver, who married Samuel Sneyd, and THESE ARE THE FIRST SIBLINGS SHE HAS EVER HAD HERE.
• AND A ONE-DIGIT DISAGREEMENT THAT NOBODY CAN SETTLE FROM HERE. The printed transcript reads «Elizabeth, d. of David Oliver, Cross St., & Mary, B. 7»; FindMyPast's index of the same register line says she was BORN 17 MAY. Two transcriptions of one entry, agreeing on the baptism date, the parents and the year, and differing by a single digit on the day of birth. NEITHER IS THE REGISTER. It is recorded as a disagreement rather than resolved by preferring whichever source was read second.
• THE SNEYD MARRIAGE IS NOT IN THE PRINTED OSWESTRY REGISTER AND THAT PROVES NOTHING, WHICH IS THE WHOLE POINT. The eight volumes print 581 Oswestry marriages and the only Sneyds anywhere in them are the GENTRY SNEYDS OF BISHTON — the same Staffordshire family the tree once tried to graft this line onto. BUT THE PRINTED MARRIAGE RUN ENDS IN 1808, and Samuel Sneyd's marriage to Elizabeth Oliver is placed about 1809. THE TRANSCRIPT STOPS ONE YEAR SHORT OF THE QUESTION. An hour was nearly spent reporting an absence that covers everything except the year that matters.
• THE WAKEFIELDS TURN OUT TO HAVE BEEN PUBLIC MEN. JABEZ WAKEFIELD stands in the obituary list of Slater's Queensland Almanac for 1904, under the 7th of February, as «AN OLD RESIDENT AND EX-MAYOR OF ROCKHAMPTON» — this archive already held his death as 8 February 1903 from the Queensland register, so the two agree to a day, and WHAT IS NEW IS THAT HE HAD BEEN MAYOR. HIRAM WAKEFIELD is a Brisbane figure across twenty years: the Aldine History of Queensland of 1888 twice describes men who «entered the employ of HIRAM WAKEFIELD, TIMBER MERCHANT»; in 1868 Slater's has him running a TWO-HORSE COACH between Brisbane and Sandgate; and from 1882 to 1891 he is a trustee and then a director, «Hiram Wakefield, Esq., J.P.», beside Hon. James Swan M.L.C. and JAMES R. DICKSON, LATER PREMIER OF QUEENSLAND.
• ONE READING WAS REFUSED. The 1888 almanac prints him «M.L.A.» where the volumes for 1882, 1883, 1884, 1885, 1886, 1887, 1890 and 1891 all print «J.P.». ONE AGAINST EIGHT IS A COMPOSITOR'S SLIP AND NOT A SEAT IN THE LEGISLATIVE ASSEMBLY, and it goes down as a discrepancy. THE SANIGARS GAVE NOTHING: «Hannah Saniger» returns four hits and all four are four scans of one Toronto history naming a different Hannah Saniger who married a George Dale. Written down because the name is rare enough that the next person will find it too.
• AND THE SNEYD ORIGIN HAS A PRINTED REGISTER AFTER ALL, THOUGH IT WILL NOT OPEN. The Staffordshire Parish Register Society printed MADELEY, and the volume is on the Internet Archive — LENDING-RESTRICTED, text returning HTTP 401, a borrow needing an account this archive will not make. ITS PUBLIC SNIPPETS STILL GAVE SOMETHING: «13 MARGARET D. OF WILLIAM SNEYD, WEAVER» and «THOMAS S. OF WILLIAM SNEYD OF GT. MADELEY». THE WORD «WEAVER» IS THE MATCH — this archive holds Samuel Sneyd baptised at Madeley All Saints on 4 June 1781, father WILLIAM SNEYD, WEAVER. WHAT IT DOES NOT ESTABLISH IS ANY RELATIONSHIP: the snippets carry NO YEARS, so a Margaret and a Thomas are children of a weaver of that name in that parish and nothing more until somebody reads the book.
• AND IT SETTLED ONE THING FOR NOTHING. THE REGISTER DISTINGUISHES «GT. MADELEY» FROM «L. MADELEY» — GREAT AND LITTLE MADELEY, two townships. This archive has written «Madeley All Saints» on the Sneyd origin from the beginning and HAS NEVER ASKED WHICH HALF THE FAMILY LIVED IN. It is the same shape as the two Ingouvilles that cost a day this week and the two Kingswoods that nearly cost the Sanigars their county — CAUGHT THIS TIME BEFORE IT COST ANYTHING, which is the first time that has happened.
• AND THE RESEARCH GOT A PAGE INSTEAD OF A REGISTER LINE. George Abbas Kooli D'Arcy had been going into this archive the way everything does — record lines in the register, rows on the search log, a paragraph in the changes — and that is right for a baptism and wrong for a life. HE NOW HAS A DOSSIER AT /ABBAS-KOOLI: the corrected name of 1837, the adjutancy, the captaincy printed twice, Bombay, the West India Regiment, the Gambia from 1859, Baddiboo and Sabba and Tubabecolong, the sword of honour, the Falklands from 1870, and the death in Devon. It is linked from his father's page, so the published gate counts it rather than calling it an orphan, and it carries the two things a page like this should carry: WHAT IS STILL NOT KNOWN (whether he ever married) and HOW HE WAS FOUND, which is the transferable part.
• AND A NEW QUESTION THIS ARCHIVE HAD NEVER ASKED. check_cites.py asks whether a citation STILL RESOLVES TODAY. IT DOES NOT ASK WHETHER THE READER COULD STILL SEE WHAT THIS ARCHIVE SAW IF IT STOPPED. tools/snapshot.py asks that: for every citation this archive CHOSE — the GEDCOM's myheritage links are somebody else's footnotes — does a capture exist in the Wayback Machine? THE FIRST ANSWER WAS 15 OF 83. A footnote to a page that has gone is a footnote to nothing, and the reader cannot tell it apart from a good one.
• THE FIRST VERSION OF THAT TOOL WAS WRONG IN A FAMILIAR WAY AND WAS CAUGHT BY RUNNING IT. It asked the Wayback Machine to capture twelve archive.org URLs AND GOT TWELVE 523s BACK — the Internet Archive quite properly declining to snapshot itself. A CITATION TO THE INTERNET ARCHIVE IS ALREADY A PERMANENT COPY, and counting those as unprotected both flattered the problem and would have sent the next person chasing a capture that cannot exist. With that fixed the honest figure was 46 OF 83, twenty-three of them permanent because they ARE the archive. Captures have since been requested for the rest, twenty at a time with a pause between, because a request to a public archive to crawl a public page should look like a request and not a flood.
• AND THE OSWESTRY VOLUMES WERE ALREADY ON DISK, SO ONE MORE QUESTION COST NOTHING. The baptisms give Elizabeth Margaret Oliver's mother only as «Mary». The printed LLANYBLODWEL register has «1772, JUNE 12. DAVID OLIVER OF OSWESTRY P: BAC: & MARY OWENS SP: … lic: mar». WHAT FITS: he is of Oswestry parish, where the baptisms are; the forename pair matches; 1772 sits properly before Dianna in 1777, Elizabeth in 1782 and David in 1789; and a David Oliver, cooper, buried at Oswestry in 1808 aged 59 was about twenty-three in 1772. WHAT DOES NOT: NOTHING LINKS THE ENTRY TO THE BAPTISMS EXCEPT THOSE FITS, and David and Mary are two of the commonest names on this border — the same eight volumes hold David Olivers at Blodwell, Trefyrclawdd, Pentreshanel, Sweeney and Willow Street across a century.
• THE CONTROL IS THAT THE SHORTLIST IS ONE LONG. All eight volumes were searched for EVERY marriage of a David Oliver: between 1765 and 1782 there is EXACTLY ONE, and the only other in the series is «1799, May 10. David Oliver, bac: & Anne Morris, sp:» — wrong wife, wrong decade. IF IT IS THE RIGHT MARRIAGE, ELIZABETH MARGARET OLIVER'S MOTHER WAS MARY OWENS and the Sneyd line gains a surname it has never had. IT IS NOT YET THE RIGHT MARRIAGE, and the register line says so in those words rather than quietly promoting a good fit into a fact.
• AND THE SNAPSHOT COUNT FINISHED AT 70 OF 84, FROM 15. The captures were requested twenty at a time with a pause between, because asking a public archive to crawl a public page should look like a request and not a flood. THE FOURTEEN THAT REMAIN ARE NOT A JOB UNDONE AND THE TOOL NOW SAYS SO IN ITS OWN VOICE. Every one of them resolves perfectly and simply declines to be archived: NLA.GOV.AU SIX TIMES — Trove's article permalinks — FINDAGRAVE THREE TIMES, FamilySearch's tree, the Anubis-gated Seine-Maritime viewer, FIBIS, and two dynamic download endpoints. A HOST THAT DECLINES TO BE ARCHIVED IS A FACT ABOUT THE HOST, and a bare number would have read as work nobody had got round to, so the tool prints them grouped by host with that sentence attached.
• AND PORTER'S THIRD CAMPAIGN TURNED OUT TO BE THE BEST OF THEM. The index gave «Copenhagen, 243» and the first full-text pass had found nothing there, so it was written down as THE INDEX'S CLAIM rather than a passage read. It is a passage, and it reads: the force sailed from Yarmouth Roads on 27 July 1807 with 20,000 men «under the command of MAJOR-GENERAL SIR ARTHUR WELLESLEY», and «THE ENGINEERS CONSISTED OF LIEUTENANT-COLONEL R. D'ARCY, COMMANDING ENGINEER; Captains R. Fletcher, J. F. Birch and S. R. Chapman; Second Captains C. W. PASLEY, H. Goldfinch and C. S. Rhodes; First Lieutenant G. Collyer.» TWO OF THE MEN UNDER HIM BECAME FAMOUS — Richard Fletcher, Wellington's chief engineer in the Peninsula, killed at San Sebastian; and Charles Pasley, who founded the school of military engineering AT CHATHAM, WHERE D'ARCY HIMSELF DIED AS COMMANDANT TWENTY YEARS LATER.
• AND THE BOOK'S OWN INDEX IS WHAT FOUND IT, WHICH IS THE LESSON. The first pass matched SEVEN instances of the surname and MISSED THIS ONE, because the scanner had set it «D‘Arcy» with a LEFT single quotation mark where every other instance in the volume uses a right one. THE SEARCH WAS WRONG AND THE INDEX WAS RIGHT. Nobody would have gone looking for page 243 if Porter had not listed it. AN INDEX IS A CONTROL ON A FULL-TEXT SEARCH — an independent statement of what the book contains, made by somebody who could read it — and after a week of full-text searching this archive now has a reason to read the index first.
• AND A TEMPTING INFERENCE WAS REFUSED. ALLEN'S INDIAN MAIL CARRIED THE RUCK MARRIAGE: «Groom RUCK, Arthur Ashley · Bride D'ARCY, ELIZABETH ELEANOR · 1877» — a second, independent record of the Penzance marriage, reported from the India side. IT IS NOT EVIDENCE THAT SHE WAS ONE OF THE INDIAN D'ARCYS, and it would have been very easy to say that it was. Arthur Ashley Ruck was himself an India-serving officer in the Afghan war of 1878-81, so the Anglo-Indian paper had every reason to print his marriage whoever he married. FIBIS holds NO BAPTISM for her across two million names. Whose daughter she was is exactly as open as it was yesterday.
• AND THE PERSIAN SERVICE STOPPED BEING A SENTENCE IN AN OBITUARY. This archive has had Joseph D'Arcy's five years in Persia only from his own death notice of 1848, looking back forty years. JAMES MORIER WAS SECRETARY OF THE BRITISH EMBASSY AND SAILED WITH HIM, and his book names him three times. The voyage out: «On board the CHICHESTER STORE-SHIP, which accompanied us, were MAJOR (NOW COLONEL) D'ARCY, and MAJOR STONE, OF THE ARTILLERY, WHO DIED IN PERSIA; and TEN PRIVATES OF THE ARTILLERY.» And then, when Sir Gore Ouseley scattered his suite across Persia to record antiquities — Sir William Ouseley to Fasa for the tomb of Cyrus, the Hon. Mr Gordon to Shouster for ancient Susa, Morier himself to Persepolis — «COLONEL D'ARCY PROCEEDED TO FIROUZABAD, WHERE WE HAD HEARD OF SOME REMARKABLE SCULPTURES.» An artillery officer sent to draw rock reliefs, which is not how this archive had pictured him.
• AND HE COMMANDED AN ATTACK, WHICH IS THE ONLY ACTION OF HIS THIS ARCHIVE HAS. Abbas Mirza marched on Mogan; eight hundred Russians with two guns held the village of SULTANBOOT. «THE ATTACK, UNDER THE COMMAND OF COLONEL D'ARCY, TOOK PLACE IN THE MORNING. The Russians, in negligent confidence, were surprised to see A WELL-DIRECTED FIRE OF GRAPE open upon them, WHICH IN A SHORT TIME KILLED 300 OF THEIR MEN… they capitulated. ONE OF THE ARTICLES OF CAPITULATION WAS, THAT THEIR HEADS WERE NOT TO BE CUT OFF.» The Persians first reported two thousand killed and five thousand taken, and told Morier to his face that they had inflated it because «THIS IS THE FIRST TIME OUR TROOPS HAVE MADE ANY STAND AT ALL AGAINST THE RUSSIANS». NO YEAR IS PUT ON IT HERE: the passage carries none, thirty thousand characters before it hold no date, and the book spans 1810 to 1816.
• AND THE INDEX FOUND HIM, FOR THE SECOND TIME IN ONE DAY. A full-text search of Morier matched ONE instance of the surname — THE INDEX ENTRY, «D'Arcy, Colonel, 2. 68. 186». All three body passages were missed because the book sets his name in LOWER CASE, «d'Arcy», and the pattern demanded a capital D. Earlier the same day Porter's Copenhagen page was missed over a LEFT rather than a right quotation mark. TWO DIFFERENT CHARACTERS, TWO MISSED PAGES, AND THE INDEX RIGHT BOTH TIMES. Porter's volume II was re-scanned with the corrected pattern in case it had hidden anything: it had not, and its nothing is now a measured nothing.
• AND THE PERSIAN SERVICE TURNED OUT TO BE A WHOLE LIFE. SIR WILLIAM OUSELEY WAS THE AMBASSADOR'S BROTHER AND SHARED A HOUSE WITH HIM AT TABRIZ, and his three volumes name Joseph D'Arcy thirty-four times. HE WAS THE EMBASSY'S ARTIST. Ouseley's preface: «THE PLATES WHICH CONTRIBUTE MOST HIGHLY TO THE EMBELLISHMENT OF THIS VOLUME have been engraved from drawings made by MY ACCOMPLISHED AND AMIABLE FRIEND… Those plates are numbered II. IV. V. XI. XVIII, and XX.» They are signed «Colonel D'Arcy del.» and they are the cave temples at ELEPHANTA, the great Sasanian sculpture at SHAPUR, a nautch dance, a well near Bushehr with mount Halilah behind it, and the Ambassador's house at Rio. At the tomb called Solomon's Mother's grave Ouseley sketched the same view and then PREFERRED D'ARCY'S TO HIS OWN. He also helped take squeezes of a Pahlavi inscription, «impressing every form of the letters on paper slightly wetted for the purpose, and applied to the very stone»; and at Isfahan he drew the poet KAZEM WALEH, eighty years old and composing his own epitaph, who refused until flattered, tried on every turban he owned, and then «SAT MOST COMPLACENTLY WHILE COLONEL D'ARCY FINISHED AN ADMIRABLE LIKENESS».
• AND IT ANSWERED A QUESTION THIS PAGE HAD WRITTEN DOWN AND LEFT OPEN. The Joseph D'Arcy page said, of the Persian service: «WHAT IS STILL MISSING IS THE YEARS HE WAS THERE AND THE NAME OF THE PRINCE HE SERVED — ABBAS MIRZA IS THIS ARCHIVE'S IDENTIFICATION, NOT THE CALENDAR'S.» Ouseley names the prince outright and dates the moment. NOVEMBER 1811, AT TEHRAN: the Prince asked for British officers, «but THE KING WOULD NOT CONSENT that any of those officers should leave Tehran WITHOUT KHELAATS, OR DRESSES OF HONOUR»; they were given them and «PROCEEDED TO JOIN THE ARMY OF ABBAS MIRZA, THE PRINCE ROYAL, AT TABRIZ». By June 1812 he was «MY FRIEND MAJOR D'ARCY, WHO, AS SENIOR OFFICER, COMMANDED IN THE MILITARY DEPARTMENT» — two Russian envoys and a French officer lodged in his house, and Ouseley left for Constantinople from it on 1 July 1812. THE ARCHIVE'S GUESS IS NOW THE AMBASSADOR'S BROTHER'S STATEMENT.
• ONE DATE IS LEFT DISAGREEING AND ONE MISTAKE OF MY OWN IS WORTH RECORDING. A footnote in Ouseley's first volume dates the Firuzabad excursion to «THE SUMMER OF 1814», but the narrative places that dispersal in 1811 and Ouseley left Persia in July 1812; either the year is a slip for 1811 or it records a later visit, AND THIS ARCHIVE DOES NOT PICK ONE. And in writing the new section I OVERWROTE THE PAGE'S EXISTING «persia» BLOCK, because the key was already taken and I did not look. The PAGES gate caught it within the minute — the old block renders plain and my markdown came through raw — and the original is restored, with the new material under its own key beside it. A GATE THAT CATCHES YOU CLOBBERING YOUR OWN PAGE IS WORTH THE TROUBLE OF HAVING WRITTEN IT.

found AddedDavid Oliver, COOPER of Cross Street, Oswestry — buried 1808 aged 59, with three children baptisedAddedElizabeth Margaret Oliver has siblings for the first time: Dianna 1777 and David 1789AddedJabez Wakefield was EX-MAYOR OF ROCKHAMPTON; Hiram was a Brisbane timber merchant and J.P. for twenty yearsCheckedthe Sneyd marriage is not in the printed Oswestry register — which stops in 1808, one year shortAddedthe printed Madeley register located — William Sneyd the WEAVER is in it, and Madeley turns out to be two townshipsAdded/abbas-kooli — a page of his own for the governor, linked from his fatherAddedtools/snapshot.py — does a PERMANENT COPY exist of every source this archive cites? 47 of 83 didAddeda candidate marriage for the Oswestry Olivers — David Oliver and MARY OWENS, Llanyblodwel 1772, graded as a candidateCheckedpermanent copies — 15 of 83 at the start, 70 of 84 now, and the 14 left are hosts that refuse to be archivedAddedRobert D'Arcy was Commanding Engineer at Copenhagen in 1807, under the future Duke of WellingtonCheckedElizabeth Eleanor D'Arcy — Allen's Indian Mail carried her marriage, and that is NOT evidence about her parentageAddedJoseph D'Arcy in Persia, from an eyewitness — he sailed with the embassy, was sent to draw ruins, and commanded an attackCheckedPorter vol. II re-scanned with the corrected pattern — genuinely no D'Arcy in itAdded/joseph-darcy — the Persian service from the Ambassador's brother: he was the embassy's ARTIST, and six plates of a scholarly work are hisChecked«Abbas Mirza is this archive's identification, not the Calendar's» — answered, by a man who was there

A novelist published this family's photographs in 1972 and nobody here had asked who she was

did

• THE WORK LIST WAS EMPTY OF ANYTHING THIS ARCHIVE COULD START, SO IT WENT LOOKING. The Internet Archive's full text was searched for «ABBAS KOOLI» — a name chosen because nothing else in print shares it — and returned 329 hits. THE FIRST IS A BOOK ABOUT THIS FAMILY. Ancestral Voices, by the novelist BERTA RUCK (1878-1978), London, Hutchinson, 1972. Its list of illustrations, which the item page shows without borrowing, is a roll-call: «Major-General Robert D'Arcy, R.E., C.B.», «Jean D'Arcy, nee Ward», «Colonel Joseph D'Arcy, R.A.», «Lady Catherine D'Arcy, nee West», «The Three Aunts D'Arcy, JANE, KATE AND CHARLOTTE», one of those aunts «in Persian dress BROUGHT HOME BY HER BROTHER JOSEPH», «George Abbas Kooli D'Arcy», «Robert West D'Arcy», «Annie Hewish D'Arcy, nee Adams» — and then «Amy Darwin, nee Ruck» and «Charles Darwin». PHOTOGRAPHS OF NINE OF THESE PEOPLE WERE PRINTED IN 1972.
• AND IT HAS BEEN HIDING UNDER A TYPING ERROR FOR FIFTY YEARS. The Internet Archive catalogues it under the subject «DORSEY FAMILY», which is a cataloguer's mangling of D'Arcy, so no search for the surname has ever surfaced it. The GEDCOM notes this archive inherited quote «by Berta Ruck» in two places and NOBODY HAD ASKED WHO SHE WAS. The answer took one search: FreeBMD gives «RUCK, Arthur Ashley» and «D'ARCY, Elizabeth Eleanor» AT THE SAME VOLUME AND PAGE — Penzance, September quarter 1877, GRO 5c 381, in a quarter marked >99% transcribed. Berta was born the following year. The tree asserted a Ruck connection; THIS IS THE FIRST RECORD OF IT.
• THE BOOK ALSO SETTLES A NAME THE TREE COULD NOT DECIDE. The GEDCOM hedged «Anne Heevish or Hewish Adams» for the wife Robert West D'Arcy married at Bombay in 1846. THREE RECORDS NOW AGREE AND NONE OF THEM COPIED THE OTHERS: the India Office ecclesiastical return indexes her ANNE H ADAMS, the Gentleman's Magazine announcement calls her ANNIE ADAMS, and her great-niece's plate list names her ANNIE HEWISH D'ARCY, NEE ADAMS. An initial in a government index, a forename in a magazine, and a caption under a photograph, and between them a woman has her whole name back.
• WHAT IS NOT CLAIMED. That Berta Ruck is the daughter of that 1877 marriage is AN INFERENCE FROM A DATE AND A FATHER'S NAME, not a record, and whose daughter Elizabeth Eleanor was is not established at all. The «C.B.» after Robert D'Arcy's name is a caption written by a descendant in 1972, not an official list, and is recorded as a lead to test against the Gazette rather than an honour this archive has documented. The book's text is in copyright and lending-restricted: ONLY THE PLATE LIST WAS READ, BORROWING NEEDS AN ACCOUNT AND NONE WAS MADE, and reading it properly is now on the work list as something a library copy would solve.
• ONE SEARCH WAS RUN THAT COULD NEVER HAVE ANSWERED, AND IS WRITTEN DOWN FOR THAT REASON. Births to a Ruck father with a D'Arcy mother, 1878-1900, returned nothing — because THE GRO BIRTH INDEX DOES NOT CARRY THE MOTHER'S MAIDEN NAME UNTIL SEPTEMBER 1911. The zero is the shape of the index and not a fact about anybody. After two days of finding instruments that answer a different question from the one asked, this archive can at least recognise one of its own.
• AND THE SAME SEARCH REACHED THE OLDEST QUESTION ON THESE PAGES. The Internet Archive's full text carries NOTES AND QUERIES for 1 DECEMBER 1866, page 432, where a correspondent asks «who was the grandfather of MAJOR-GENERAL ROBERT D'ARCY, R.E., WHO DIED COMMANDANT OF CHATHAM IN 1827?» and prints, as FAMILY TRADITION, the answer this archive has spent months disputing: «Colonel d'Arcy, A BRANCH OF THE HOLDERNESS FAMILY FROM YORKSHIRE, commanded a regiment during the American War. He was KILLED IN ACTION, leaving two sons: 1. Robert, Major-General R.E. 2. CONSTANTINE, Lieut. R.A., DIED S.P. On his voyage to America Colonel d'Arcy and his family were WRECKED. They lost all their baggage and papers, which, it is believed, would have proved their noble descent.»
• THREE THINGS MAKE THAT WORTH MORE THAN THE STORY IT CARRIES. It is ONE HUNDRED AND FIFTY-NINE YEARS OLDER than the family tree that states Yorkshire as a fact, so the tree did not invent it and nor did MyHeritage. The man who printed it was ASKING, NOT ASSERTING — he wanted the grandfather and did not have him. And THE EXCUSE FOR THE MISSING EVIDENCE IS INSIDE THE STORY: a shipwreck that drowned exactly the papers which would have proved the descent, which is the shape a pedigree takes when it cannot be documented. A full-text search for «Commandant of Chatham» returns 144 hits across everything the Internet Archive holds and the only one in Notes and Queries is the question itself. NOBODY EVER ANSWERED IT. The question has stood open in print since 1866 and this archive has now joined the queue.
• IT ALSO HANDS OVER THE FIRST TESTABLE THING THE TRADITION HAS EVER OFFERED. A man who COMMANDED A BRITISH REGIMENT AND DIED IN ACTION between 1775 and 1783 is not obscure: he is in the Army Lists, his death is gazetted and his successor is gazetted after him. A first pass found nine «Arcy» notices in the Gazette for 1775-1790 and the officers in them are a Brevet Major FRANCIS D'Arcy and an Ensign JOHN POMEROY D'Arcy — no colonel. THAT IS NOT A DISPROOF AND IS NOT FILED AS ONE: this archive has watched the same Gazette OCR turn the surname into «D'Aroy» inside a single career, and the site will not page past its first ten results. The Army Lists themselves are free and full-text on the Internet Archive, and the work list now says so.
• AND THE GOVERNOR GOVERNED TWICE. A Falkland Islands bibliography and Henige's Colonial Governors between them give «Colonel George Abbas Kooli D'Arcy: GOVERNOR OF THE FALKLAND ISLANDS 1870-1876», after the Gambia from 1859 — two colonies and the Atlantic between them, and this archive knew of neither until this week. The same search finally identified the «military records for 1840 and 1841» that have sat on the work list as a bare phrase: they are the MONTHLY ARMY LISTS OF 1840, which print «Adj. Geo. Abbas Kooli [D']Arcy» and confirm he was ADJUTANT of the 94th. Notes and Queries adds, in 1866, that he was then «THE PRESENT REPRESENTATIVE OF THE FAMILY», claiming descent from Edward III THROUGH HIS MOTHER — Catherine Georgiana West — with arms and a crest. A claim reported, not a pedigree proved, and graded accordingly.
• AND THEN THE TRADITION WAS TESTED, THE SAME DAY IT WAS FOUND, AND IT SPLIT CLEANLY IN TWO. Four official Army Lists — 1775, 1776, 1778 and 1783, lists of GENERAL AND FIELD OFFICERS — were downloaded and searched entire. THE TWO SONS ARE BOTH THERE. The 1783 volume puts ROBERT D'ARCEY in THE CORPS OF ENGINEERS, among the lieutenants, in a block of officers all dated 17 JANUARY 1776 — THE EXACT DATE THIS ARCHIVE ALREADY HELD for his first commission, printed seven years later and reached by a different road. And it puts CONSTANT[INE] D'ARCEY, 12 July, among the FIRST LIEUTENANTS OF THE ROYAL REGIMENT OF ARTILLERY, which is word for word what the tradition claimed: «Constantine, Lieut. R.A.»
• THE COLONEL IS IN NONE OF THEM. No Colonel D'Arcy appears in 1775, 1776, 1778 or 1783. The only officers of the surname are Francis D'Arcy, John Pomeroy D'Arcy, and a George and a John in the indexes. THE CONTROL IS THE VOLUMES THEMSELVES: they use the word «Colonel» 230, 235, 245 and 312 times, so they are not silent about colonels. THREE GAPS KEEP IT SHORT OF A DISPROOF and are stated rather than buried — a provincial or loyalist colonel need not be listed, nor one on the Irish establishment, and «Colonel» may have been a local or courtesy rank. BUT THE SHAPE IS HARD TO MISS. A tradition recorded in 1866 is right about the two men it names who can be checked, and silent of support for the one man it needs. THE SHIPWRECK THAT DROWNED THE PAPERS WAS DOING A LOT OF WORK IN 1866 AND IT IS DOING IT STILL.
• AND THE DEPARTMENT THAT APPOINTED HIM HAD WRITTEN HIS LIFE DOWN ALL ALONG. The Colonial Office List for 1877 is free, out of copyright, and its whole text is a single 2.2 MB file. Its biographical notice runs: adjutant of the 94th for three years, SIXTEEN YEARS IN THAT CORPS; AUTHOR OF A WORK ON «LIGHT INFANTRY DRILL»; commanded the 3rd West India Regiment for several years; served in the SOUTHERN MAHRATTA AND SAWANT-WARREE CAMPAIGN OF 1845-6 UNDER BRIGADIER OUTRAM; AIDE-DE-CAMP AND ASSISTANT PRIVATE SECRETARY TO FIVE SUCCESSIVE GOVERNORS OF BOMBAY and to Sir Willoughby Cotton commanding the Bombay army; SOLD OUT IN 1858; Governor and Commander-in-Chief of the Gambia from JUNE 1859; an expedition against the robber-king of BADDIBOO that earned him the thanks of the Secretary of State; the BATTLE OF SABBA, 21 FEBRUARY 1861.
• AND A VICTORIA CROSS WAS WON FOLLOWING HIM THROUGH A STOCKADE. At Tubabecolong in 1866 he embarked at FORTY-EIGHT HOURS' NOTICE against a Marabout invasion and carried the entrenched position. TWO OFFICERS WERE KILLED AT HIS SIDE — Lieutenants Kelly and Jenkins of the 4th West India Regiment — with four soldiers, two volunteers, four of the native contingent and sixty-one of all ranks badly wounded. «PRIVATE SAMUEL HODGE, WHO FOLLOWED COLONEL D'ARCY THROUGH THE STOCKADE, RECEIVED THE VICTORIA CROSS.» Hodge was among the first black soldiers to receive it. On leaving the Gambia in January 1867 the European merchants gave him a sword of honour worth £120. ONE EDITORIAL CHOICE IS ON THE RECORD: the 1877 entry describes the invaders in the racial language of its period and that phrase is not reproduced here, because nothing in this family's history turns on it.
• AND IT QUIETLY EXPLAINS THE BOMBAY MARRIAGE. George was aide-de-camp to five governors of Bombay and campaigning in the Southern Mahratta country in 1845-6. HIS BROTHER ROBERT WEST MARRIED ANNIE HEWISH ADAMS AT BOMBAY ON 20 JANUARY 1846. That is an observation and it is filed as one: NO SOURCE PLACES THE TWO BROTHERS IN THE SAME ROOM, only in the same presidency in the same months. But the archive now knows why there were D'Arcys in Bombay at all, which it did not on Thursday.
• AND THE BOOK THIS ARCHIVE HAS BEEN CITING FOR MONTHS HAD NEVER BEEN READ. Porter's History of the Corps of Royal Engineers was in the sources as «Porter's station list», the authority for Robert D'Arcy being posted to Barbados. ITS INDEX GIVES HIM THREE CAMPAIGNS: «D'ARCY, Robert, at second siege of MINORCA, 209, 216; COPENHAGEN, 243; WALCHEREN, 248.» At MINORCA he is one of six Engineer officers inside Fort St Philip's when the Duc de Crillon landed on 19 August 1781 — a garrison of three thousand, besieged six months. At WALCHEREN in 1809 he is something else entirely: of five corps in an army of forty thousand under the Earl of Chatham, the largest was Sir Eyre Coote's, and «THE COMMANDING ROYAL ENGINEER WITH THIS CORPS WAS LIEUTENANT-COLONEL R. D'ARCY». JOSEPH'S OBITUARY, HELD HERE SINCE SEPTEMBER, SAYS JOSEPH WAS AT THE TAKING OF WALCHEREN TOO — the father commanding the Engineers of the main corps, the son a Royal Artillery officer of twenty-nine, and neither record knows about the other.
• HE WAS ALSO PUT UNDER ARREST THERE, AND THE REASON IS THE ONLY PIECE OF HIS CHARACTER THIS ARCHIVE HAS. Colonel Fyers, shot through the chest at Flushing, ordered the senior engineer in the trenches to hand in a written report every morning. The Brigade-Major wrote it down: «The first morning, Colonel D., AN OLD MAN BEYOND SIXTY, refused compliance, saying HE HAD OTHER THINGS TO DO, and no solicitation could prevail on him to write a single line.» The third morning he refused again, «SAYING IT WAS ALL NONSENSE», and denied his superior's right to ask him for anything «BECAUSE SOME THIRTY OR FORTY YEARS PREVIOUSLY HE HAD BEEN SENIOR TO COLONEL F.» He was arrested; the preliminaries of a court-martial were gone through; and then he «thought better of it, made his apologies, was released, and ever after furnished his report». Porter names him only in a footnote.
• AND THE INSULT CARRIES A DATE. «AN OLD MAN BEYOND SIXTY» IN AUGUST 1809 PUTS HIS BIRTH BEFORE ABOUT 1749. The family tree says 1751, which would have made him fifty-eight that summer and not «beyond sixty». IT IS AN EYEWITNESS IMPRESSION PRINTED EIGHTY YEARS LATER AND NOT A REGISTER, and it is graded as that. But the tree's 1751 is itself only arithmetic from an age on a burial register, so this is now TWO SOFT NUMBERS DISAGREEING rather than one standing unopposed — and it is the sort of thing /hornby exists to keep track of.
• AND THE SAME TRICK WAS TURNED ON THE OLDEST QUESTION IN THE ENGLISH LINE. Item 106 has been asking for THE BERKELEY REGISTER ITSELF since September, because every index has failed. PHILLIMORE'S «GLOUCESTERSHIRE PARISH REGISTERS: MARRIAGES» IS A PRINTED TRANSCRIPT SERIES, free and out of copyright, and A TRANSCRIPT IS THE REGISTER. Six volumes were downloaded whole and read. They print Ampney Crucis, Naunton, Prestbury, Sevenhampton, Thornbury, Kingscote, Painswick, Batsford, Beverston, Ebrington, Elkstone, Standish, Stinchcombe, Weston Birt, Willersey, Stone and Nether Swell. BERKELEY IS NOT AMONG THEM — every «Berkeley» in the text is a bride or groom described as «of Berkeley» in somebody else's parish. No John Saniger marries any Lydia in any of the six, against a control of THIRTY-SIX SANIGAR MARRIAGES PRINTED ACROSS THE SERIES, so the surname is in no way invisible to it.
• ONE THING FELL OUT OF IT THAT NO PAGE HERE HAS EVER SAID. THE SURNAME IS THICKEST AT STONE, the parish adjoining Berkeley — twenty of the thirty-six entries, in the 1670s to 1690s, with the rest at Thornbury, Cam and Dursley. This archive's Sanigars are at Berkeley and Ham and it has never asked where the name itself lives. IT IS DISTRIBUTION AND NOT DESCENT, and is recorded as that: not one of the thirty-six is shown to be a relative of anybody here.

found AddedBerta Ruck's Ancestral Voices (1972) — portraits of nine of these people, catalogued under «Dorsey family»AddedElizabeth Eleanor D'Arcy married Arthur Ashley Ruck, Penzance, September quarter 1877, GRO 5c 381AddedAnnie Adams's middle name settled at HEWISH by three records that did not copy each otherAdded/hornby — the Yorkshire descent in print in 1866, called «family tradition» by the man asking for proof of itAddedGeorge Abbas Kooli D'Arcy governed the Falkland Islands 1870-1876 as well as the GambiaCheckedthe Army Lists of 1775-1783 — Robert and Constantine are in them, the colonel who was to prove the descent is notAddedthe Colonial Office's own biography of George Abbas Kooli D'Arcy — a book he wrote, a battle, and a Victoria Cross won beside himAddedRobert D'Arcy at Minorca 1781, Copenhagen and Walcheren 1809 — and under arrest at FlushingCheckedPhillimore's printed Gloucestershire marriages, six volumes — Berkeley is not in the series, and the Sanigars are thickest at Stone

A bride this archive could not afford to see, named twice over for nothing

did

• OWN ERROR 56, AND IT IS THE FOURTH OF ITS EXACT SHAPE IN TWO DAYS. Yesterday this archive published that TWELVE citations no longer lead anywhere, «four of them FamilySearch collection URLs this archive chose itself, which makes them the worst four on the list». ALL FOUR ARE ALIVE AND ALWAYS WERE. urllib sends no Accept header unless it is told to, and FamilySearch answers a request without one with a flat 404. The rewrite of check_cites.py the day before had already caught three faults of precisely this kind — a swallowed backslash, a 302 graded as a refusal, a wrapped DNS error — and then shipped a fourth, because the tool was tested against what it reported rather than against what a reader sees. OPENING ONE OF THE DEAD ONES IN A BROWSER TOOK UNDER A MINUTE AND WAS NOT DONE BEFORE PUBLISHING.
• WITH THE HEADER SENT THE PICTURE IS BETTER THAN THE ONE THAT WAS PUBLISHED, AND THE GOOD NEWS IS THE PART THAT MATTERS. 300 RESOLVE, 264 WALLED, 6 TIMED OUT, 4 HOSTS GONE, 5 REFUSED. NINE candidates for a dead citation, and EVERY ONE OF THE NINE WAS INHERITED FROM THE GEDCOM EXPORT — they appear only in families.json, ancestors.json and line.json, which are built from the MyHeritage file. NOT ONE CITATION THIS ARCHIVE CHOSE FOR ITSELF IS DEAD. The nine are three links into the AIF Project and one to Brisbane's grave search, neither hostname resolving any more; two to a personal genealogy site; a Familypedia page on the old wikia.com domain; an Australharmony page at Sydney; and an EMPTY WAR MEMORIAL SEARCH URL, which is a search box somebody saved rather than a citation at all.
• AND THE TOOL NOW DRAWS THAT LINE ITSELF RATHER THAN LEAVING IT TO WHOEVER READS THE OUTPUT. check_cites.py marks every dead link «OURS» or «inherited» and closes with «of those, 0 are citations THIS ARCHIVE CHOSE and 9 were inherited from the GEDCOM export. Only the first number is anybody's job here.» NOTHING IS BEING REPAIRED AND NOTHING DELETED: mending another person's footnote inside an inherited file would make it look like this archive's work and would be overwritten by the next export.
• THE MARRIAGE HAS SAT HERE SINCE AUGUST AS A BARE FACT WITH THE WORDS «TRANSCRIPT PAYWALLED EVEN ON A PAID TRIAL». Robert West D'Arcy, second son of Joseph, married at Bombay in 1846 and that was all this archive could say. FIBIS — the FAMILIES IN BRITISH INDIA SOCIETY database, free, indexing India Office material, and never opened here — names her twice from two separate indexes. The Gentleman's Magazine marriage announcements: «ROBERT WEST D'ARCY × ANNIE ADAMS, 20 JANUARY 1846, BOMBAY». The Index of Ecclesiastical returns, IOR BOMBAY MARRIAGES: «Husband ROBT WEST D'ARCY · Wife ANNE H ADAMS, 1846». TWO INDEXES, ONE MARRIAGE, AND HER NAME SPELLED TWO WAYS BY PEOPLE WHO WERE NOT COPYING EACH OTHER. The day comes from the magazine; the India Office return gives only the year.
• AND THE ANGLO-INDIAN PRESS FOLLOWED HIM HOME TO WALES. Allen's Indian Mail printed the Bombay marriage twice in 1846 — once as «Robert W.», once as «Robert West» — and then, sixteen years later, «D'ARCY, Robert · BANGOR, N.WALES · 1862». This archive already had the English notice of that death, «late of the Honourable East India Company's Army, dead at Bangor on 14 June». Now the India side of his life records the same death from the other direction, which is what ties the Bangor man to the Company officer without anybody having to assume it.
• WHAT FIBIS DID NOT GIVE IS THE THING THE ERRAND WAS OPENED FOR. There is no cadet paper, no commission, no medal roll entry and no child. The India Office papers at the British Library remain the document that would give birth, parentage and career together, and nothing here replaces them.
• AND THE COUNT THAT LOOKED LIKE A HARVEST WAS MOSTLY GHOSTS. A default FIBIS search for «Robert D'Arcy» returns TWELVE SOURCES, which reads like twelve findings. THE DEFAULT IS SOUNDEX, AND ON THIS SURNAME IT MATCHES TRACY, TRACEY, TRAISE, DRUCE AND D'ROSE. The «Gentleman's Magazine death announcement for Robert D'Arcy» is Robert Alexander DRUCE, dead in 1829. The «Register of European Soldiers, Bengal Army» entry is Robert TRACY, gunner. Both Indian Mutiny Medal Roll hits are Tracey and Traise. The five «CMB» entries are a D'Rose family at Lucknow in the 1920s. TEN OF TWELVE. A RESULT COUNT FROM THIS DATABASE IS NOT A COUNT OF ANYTHING UNTIL THE ROWS ARE OPENED, and the same search set to exact returns 29 sources for D'Arcy in the whole database, which is the figure worth quoting.
• ONE MORE ROUTE TO LYDIA CLOSED, AND CLOSED PROPERLY. Every earlier attempt at her maiden name went through FindMyPast. FREEREG IS A SEPARATE TRANSCRIPTION PROJECT WITH SEPARATE COVERAGE AND HAD NEVER BEEN ASKED. It cannot answer either, and the three parishes that matter fail in three different ways: BERKELEY holds 2,916 records and the whole run is 1600–1677, stopping a hundred and twenty years short; WOTTON-UNDER-EDGE, where she was born, holds 16,105 and STARTS AT 1813; THORNBURY spans the 1790s and looked like the way in. A search of Thornbury AND UP TO A HUNDRED NEARBY PARISHES for any Lydia marrying 1790–1805 returns nothing — AND SO DOES THE SAME SEARCH FOR ANY MARY. That control is the whole value of the exercise: zero Marys marrying across a hundred parishes in sixteen years is not a fact about Gloucestershire, it is the absence of transcribed marriages, and without it the Lydia zero would have looked like evidence.
• AND WHETHER THE GOVERNOR OF THE GAMBIA EVER MARRIED IS ASKED AND NOT ANSWERED. FreeBMD holds FOUR George D'Arcy marriages in England and Wales across 1837–1890 — Norwich 1864, Westminster 1884, Hackney 1888, West Ham 1889 — in quarters FreeBMD itself marks >99% transcribed, and NONE CARRIES HIS FORENAMES. The GRO index records every forename a man registered under and a «George Abbas Kooli» does not hide as a plain «George». IT IS A WEAK NULL AND IS FILED AS ONE: all four results fall in September quarters, which across four records deserves somebody's second look, and he spent most of his adult life outside English registration altogether — the 94th Foot, the 3rd West India Regiment from 1852, the Gambia from 1859. A marriage at Bathurst would not be in this index at all.
• AND THE WORK LIST WAS TELLING A FLATTERING LIE ABOUT ITSELF. Its own definitions are plain: NEXT MEANS READY TO START, BLOCKED MEANS WAITING ON A PERSON OR AN INSTITUTION. Eight items sat under «next» and five of them had not been ready to start for weeks — Barbados, which its own note said in September needed the originals; Lydia, which needs the Berkeley register or the diocesan bonds; Joseph's sons, which now need the India Office cadet papers and a Gambia register; Margaret's baptism and the Ruabon baptism, both of which need a register image after today's measurements closed the free indexes. ALL FIVE ARE NOW BLOCKED AND EACH SAYS WHAT IT IS WAITING FOR. A list that calls a wall «next» is a list that makes the archive look busier than it is.
• WHAT IS LEFT UNDER «NEXT» IS THREE ITEMS AND ONLY ONE IS THIS ARCHIVE'S. Two belong to other sessions in the estate — the Luwinski and Our Family work lists, and the shared kit's gazetteer. The third is WELSH NEWSPAPERS ONLINE, tried three times across two days and stopped every time by the same Cloudflare interstitial. IT STAYS «NEXT» DELIBERATELY RATHER THAN BEING FILED AS BLOCKED: it is free, it wants no account and no fee, nobody has to write a letter or pay anybody, and somebody with an ordinary browser will open it in one click. Nothing is waiting on a person or an institution — only on a bot check that dislikes this browser, and that is not the same kind of wall.

found Corrected/worklist — the twelve dead citations are nine, and none of them is this archive's. Own error 56AddedRobert West D'Arcy married ANNIE ADAMS at Bombay, 20 January 1846 — from two free India Office indexesAddedAllen's Indian Mail reported his death at Bangor in 1862, having printed his marriage twice in 1846CheckedFIBIS — ten of the twelve sources a default search returned were soundex ghosts, not D'ArcysChecked/worklist 106 — FreeREG cannot answer Lydia's marriage, and a MARY control proves the zero is coverageChecked/worklist 134 — four George D'Arcy marriages in the English index, none of them hisChecked/worklist — five items were sitting under «next» that nothing here can start. They are blocked, and now say so

Two deaths chased at Kew for a month are read in the French register, and one of them gives an age

did

• THE ACTS WERE READ, AND THE THING THIS PAGE HAS WANTED FOR A MONTH IS AN AGE. Margaret Jones née D'Arcy died at SIX IN THE MORNING ON 1 FEBRUARY 1843, at her house in the rue de l'Hôpital, Ingouville, AGED SIXTY-ONE — which puts her birth about 1781. It is the FIRST DATING OF A CHILD OF ROBERT D'ARCY THAT COMES FROM A RECORD rather than from a tree. Her husband died seven weeks later, on 21 MARCH, at eight in the morning in the rue du Champ de Foire, AGED SIXTY-EIGHT, born about 1774. Each act names the other person: hers calls her his wife, his calls him her widower. A marriage entered at Chatham in 1806 is confirmed in 1843 by French clerks who had never heard of Chatham and spelled both names by ear — «Maddack», «Marguéritta Maria Elisabella Darcy».
• AND NEITHER ACT CAN CARRY ROBERT D'ARCY, WHICH IS THE POINT OF THE ERRAND AND IS NOT ANSWERED. Both record the birthplace of the dead person as unknown, and his adds that the names of his father and mother are unknown too. The declarants were four neighbours — a police appariteur of twenty-two, a farmer of forty-eight, two men of private means of fifty-nine and fifty-eight — and every act notes that they were NOT RELATIVES. Two English people in their sixties had died in a French suburb among people who did not know where either of them came from. The D'Arcy parentage still rests where it rested, on the English notices. ERRAND 35 IS MARKED ANSWERED FOR THE AGE AND SAYS IN ITS OWN ENTRY THAT THE PARENTAGE IT WAS OPENED FOR IS STILL OPEN.
• OWN ERROR 55. A DÉCENNALE TABLE WAS READ RIGHT THROUGH FOR THE WRONG INGOUVILLE. Cote 4 E 10127 was picked out of the search results for being the smallest index — forty-eight views — and its 1843–1852 décès table was paged to the J section and read: nine entries, Jean, Jouette, Junais, and no Jones. The verdict was worthless. TWO THINGS ON THE SCREEN SAID SO AND NEITHER WAS READ. The image URL filed the volume under «ingouville-sur-mer», a village near Saint-Valery-en-Caux sixty kilometres up the coast. And forty-eight views cannot hold thirty years of births, marriages and deaths for a town of fifteen thousand — the right table, 4 E 9933, spends THREE HUNDRED AND NINE VIEWS ON ONE DECADE. The surnames gave it away too, had anyone asked: Jouette, Blanquet, Barthélemy. A REGISTER IS NOT IDENTIFIED BY ITS TITLE. It is identified by its provenance and its size, and both were on the page.
• THE ACT WAS FOUND BY DATE, NOT BY INDEX, WHICH IS WHY THE WRONG TABLE COST NOTHING IN THE END. The 1843 deaths register of Ingouville, 4 E 9077, runs chronologically at about two acts to a view, so a known date is a binary search: act 116 is 23 March, act 85 is 1 March, act 65 is 16 February. Jones's death on the 21st had to be act 113, and it is. Margaret's was narrowed by her own notice — the Morning Herald printed it on 13 February, so the death was days earlier — and act 43 is the first of the month. THE INDEX WAS NEVER NEEDED.
• AND THE PAGE THAT GRADES EVERYTHING ELSE HAD A NUMBER NOBODY WAS GRADING. /errands carried the heading «50 OPEN · 8 ANSWERED AND KEPT» as a hand-typed string in its data, while the rows under it held FIFTY-TWO AND FOURTEEN. The page already had a counting helper and was not using it for the heading. A count that is retyped is a claim that can go stale in silence, which is the one failure this archive builds gates against — so the eyebrow is now computed from the rows, and answering an errand cannot leave the heading behind. The same tally on /changes was already computed; its data file carried a dead stale copy, which is gone.
• THE ARCHIVE'S IMAGES WERE READ THROUGH THE BROWSER, NOT SCRIPTED. The Seine-Maritime viewer serves IIIF, and a first pass pulled page crops with curl until the server began returning placeholder tiles. The browser held a «techaro.lol-anubis-auth» cookie — a proof-of-work token earned by passing a bot check. Replaying that token outside the browser would have been working around the gate, so the scripting stopped there and the rest of the reading was done by navigating the browser to the crops, which is what the viewer itself does. THE RULE IS NOT «GET THE IMAGES»; IT IS «DO NOT DEFEAT A BOT GATE», AND IT COST NOTHING TO KEEP.
• AND THE AGE TURNED OUT TO CONFIRM SOMETHING NOBODY HAD CHECKED. This archive already carried 26 SEPTEMBER 1781, PORTSMOUTH for Margaret's birth — out of the family tree, never matched to a record, and sitting in a table beside her brother's documented one. A woman born that day is SIXTY-ONE YEARS AND FOUR MONTHS OLD on 1 February 1843, which is exactly the age the act gives. The two men who gave it were neighbours who did not know her birthplace or her parents, and said so in the same sentence. They could not have known the date and they confirmed it. It also seats her in the family: Joseph was born 14 March 1780, she is eighteen months behind him, and the 1806 notice calls her the ELDEST DAUGHTER. Three statements, none written with the others in view.
• A REDISCOVERY WAS CAUGHT BEFORE IT WAS PUBLISHED, AND IT IS WORTH RECORDING THAT IT WAS NEARLY NOT. Sweeping FreeREG for a D'Arcy daughter turned up JOSEPH D'ARCY, BAPTISED PORTSEA ST MARY 19 MARCH 1780, FATHER ROBERT, MOTHER JEAN — which reads like the first parish record naming Robert D'Arcy as a father, and was written up in this session's head as exactly that. IT IS ALREADY IN THE REGISTER, CITED TO FREEREG, with the Hampshire Genealogical Society's independent transcription beside it. The archive's own «memory» error class has ten entries and every one of them is this: announcing as new something already held. The eleventh was avoided by reading build_register.py before writing a word, which took under a minute and should not have been optional.
• WHAT THE SWEEP DID BUY IS A MEASURED FLOOR UNDER WORK LIST 142. Portsea St Mary's whole FreeREG holding is 992 RECORDS, ALL BAPTISMS, 1780–1782 — so a baptism of September 1781 would be inside it, and Hampshire returns Joseph alone on both exact and loose spellings. Portsmouth St Thomas holds 9,079 records, ALL MARRIAGES, 1651–1776, and no baptisms at any date. Old Portsmouth holds nothing. Nationally there are eighteen exact D'Arcy baptisms between 1775 and 1790 and the only Margaret among them is at Happisburgh in January 1775, which would make her sixty-eight at death rather than sixty-one. The next step is a register image, not another index.
• AND A SEARCH NOBODY SHOULD REPEAT. WORK LIST 143 WANTED A RUABON BAPTISM FOR JOHN MADDOCK JONES ABOUT 1774. FreeREG holds 30,976 transcriptions for RUABON ST MARY and the baptism column is BLANK FOR THE 1770s, 1780s, 1790s AND 1800s, starting at 1,078 in 1810–1819 — while those same 1770s carry 253 marriages and 722 burials. The original register begins in 1559; it is the transcription that stops. The Welsh-named duplicate place «Rhiwabon» in the same county list is an empty stub and not a second chance. On Welsh custom his middle name should be his mother's maiden surname, so Denbighshire was swept for MADDOCK across 1750–1785: EIGHTY-SEVEN RECORDS AND NONE AT RUABON, with ten Ruabon rows returned in the same result set to prove the parish was in scope. Three Jones–Maddock marriages exist in the county and the nearest in date, Gresford in January 1772, IS A RESEMBLANCE AND IS ATTACHED TO NOBODY.
• A NATIONAL LIBRARY THE ATLAS DID NOT HAVE. Looking for a free copy of the Caernarvon & Denbigh Herald obituary this archive cites through a paywall led to WELSH NEWSPAPERS ONLINE, free and covering 1804–1919 — and to the discovery that the Record Atlas listed NO WELSH PROVIDER AT ALL except Findmypast and the GRO, both paid. The National Library of Wales is now a row there. Its newspaper service did not open today: a Cloudflare «Performing security verification» interstitial that did not clear in about forty seconds, on the search URL and on the bare homepage. IT WAS NOT WORKED AROUND, the Library's main site loaded normally in the same session, and the access class was left «unsurveyed» rather than guessed from one browser on one day.
• AND THE LINK CHECKER WAS NOT SLOW, IT WAS WRONG. check_cites.py has existed for weeks, wired into nothing, and the reason on record was speed: 581 URLs one at a time, no output until the last returned, a run abandoned at eight minutes. That was the smaller fault. IT READ STATUS CODES AND NOTHING ELSE, and myheritage.com — 428 of the 581, carried in from the GEDCOM export rather than chosen here — ANSWERS A ROBOT WITH HTTP 200 AND AN 843-BYTE BODY READING «Request unsuccessful. Incapsula incident ID: …». Every one of them would have been counted as a healthy citation. The headline would have been confident, reassuring and wrong about three quarters of what this archive cites. A STATUS CODE IS NOT A PAGE; the rewrite reads the first four kilobytes and calls that outcome «walled», which is neither alive nor dead but is what is true.
• RUNNING IT PROPERLY FOUND THREE MORE BUGS, ALL IN THE DIRECTION OF CRYING WOLF. Ten URLs had swallowed a trailing backslash out of JSON escaping and one had swallowed an « », so «nla.gov.au/nla.news-article3635563\» came back 404 and the tool was INVENTING dead citations. 436 MyHeritage 302s were being graded «refused» and printed under a heading that invited somebody to read them as dead. And urllib.error.URLError WRAPS the name-resolution failure rather than letting socket.gaierror through, so a host that does not exist was filed beside a host that was merely slow. A report that is read is a report that does not cry wolf, and this one would have cried wolf eleven times before it said anything true.
• IT IS NOW IN THE BUILD, AT A TENTH OF A SECOND. The network sweep is a separate deliberate run of about two minutes; the build reads a committed cache, prints how old the numbers are, and says so in its own voice past sixty days — because a cache is a measurement with a date on it and goes stale without ever looking wrong. The tally that day: 105 RESOLVE, 428 REDIRECT, 26 WALLED, 6 TIMED OUT ON THIS MACHINE, 4 HOSTS GONE, 8 REFUSED — TWELVE CANDIDATES, AND THE NUMBER WAS WRONG. See 10 October: urllib sends no Accept header unless told to, four of the twelve were alive, and the real figure is NINE, none of them a citation this archive chose. OWN ERROR 56.
• SUPERSEDED THE NEXT MORNING, AND BADLY. This entry said four of the twelve dead links were FamilySearch collection URLs chosen by this archive, “which makes them the worst four on the list”. ALL FOUR WERE ALIVE. The checker was not sending an Accept header and FamilySearch answers a request without one with a flat 404. What survives is the other half: graves.brisbane.qld.gov.au and three links into the AIF Project at aif.adfa.edu.au genuinely no longer resolve as hostnames — and all of those were INHERITED FROM THE GEDCOM EXPORT too. Own error 56, on /corrections.
• AND THE LAST THING THE DAY FOUND WAS A COLONIAL GOVERNOR. Work list 134 wanted to know «which corps, which rank» for GEORGE ABBAS KOOLI D'ARCY, Joseph's eldest son, baptised at Westminster in 1818 and known to this archive only from «military records for 1840 and 1841» and a death in Devon in 1885. THE GAZETTE HAS TEN NOTICES ON HIM AND IT IS FREE AND THIS ARCHIVE ALREADY CITES IT. Ensign in the 94TH REGIMENT OF FOOT by April 1837 — the notice exists only to correct the War Office: «The Christian names of Ensign D'Arcy, of the 94th Regiment of Foot, are George Abbas Kooli». Lieutenant 1839. Captain 1846. Major of the 3RD WEST INDIA REGIMENT in 1852, «from the 94th Foot». Lieutenant-Colonel 1854. And on 17 JUNE 1859: «COLONEL GEORGE ABBAS KOOLI D'ARCY TO BE GOVERNOR AND COMMANDER-IN-CHIEF IN AND OVER HER MAJESTY'S SETTLEMENTS IN THE RIVER GAMBIA», gazetted in the same sentence as the new Governor of Hong Kong.
• THE IDENTIFICATION DOES NOT REST ON THE NAME BEING ODD, THOUGH IT IS THE ODDEST IN THIS ARCHIVE. It rests on something already on these pages: ABBAS KOOLY IS ABBAS QOLI, FOR PRINCE ABBAS MIRZA, WHOM HIS FATHER HAD SERVED FIVE YEARS IN PERSIA — which comes out of Joseph's own obituary, found weeks before anybody looked for the son's career. A boy baptised at Westminster on 31 August 1818 is eighteen when the War Office gazettes him an Ensign in the spring of 1837. The name is not a coincidence to be explained away; it is the evidence, and the thing it points at was already documented here.
• AND THE SEARCH WORKED FOR A REASON WORTH WRITING DOWN. The Gazette's OCR mangles him freely — «Qeorge», «Greorge», «D'Aroy» — so a search on the surname would have missed most of it, and a search on the forename would have drowned. KOOLI SURVIVES EVERY MANGLING BECAUSE NOTHING ELSE IN THE CORPUS LOOKS LIKE IT. Ten is a floor rather than a ceiling: an OCR slip could still be hiding a later notice, nothing after 1859 was found, and WHETHER HE EVER MARRIED IS STILL UNANSWERED. One 1857 notice lists him among officers of mixed rank each with a date; THE HEADING OVER THAT LIST WAS NOT READ, so what it conferred is not claimed — only that he is a Colonel by 1859.

found Added/margaret-jones — act N° 43, Ingouville, 1 February 1843. Margaret D'Arcy, aged SIXTY-ONE, born about 1781Added/margaret-jones — act N° 113, Ingouville, 21 March 1843. John Maddack Jones, Major, aged SIXTY-EIGHT, «veuf de Marguéritta Maria Elisabella Darcy»Corrected/searched — a décennale table was read right through for the WRONG INGOUVILLE. Own error 55Checked/errands — the heading said 50 open and 8 answered over 52 and 14. The tally is now computed, not typedClosed/errands 35 — answered by a free register, after a month of pursuing it through a paywallChecked/margaret-jones — her tree birth date of 26 September 1781 is confirmed to the month by an age two strangers gave in IngouvilleChecked/worklist 143 — FreeREG holds no Ruabon baptisms before 1810, so that search cannot be done there by anybodyAdded/searched — the Portsmouth parishes measured: one of the three can answer for 1781, and she is not in itCheckedevery outward citation, for the first time — 577 URLs in two minutes, and the checker itself was wrong about 428 of themAddedbuild.sh runs the citation report from a committed cache in a tenth of a secondAddedJoseph D'Arcy's eldest son governed the Gambia — seven Gazette notices, Ensign 1837 to Governor 1859

The research reached the pages and left four people behind, and split two more in half

did

• AND THE BEST FIND OF THE DAY CAME OUT OF A THROWAWAY LINE. The Record Atlas session mentioned eight French departmental archives that refuse a script and open to a browser. This archive has TWO DEATHS IN FRANCE — Margaret Jones née D'Arcy at Havre in February 1843 and her husband Major John Maddock Jones at Ingouville in March — and had pursued both ONLY through British consular returns at Kew, which are «miscellaneous» by their own title. NO FRENCH ARCHIVE HAD EVER BEEN TRIED.
• THE SEINE-MARITIME ARCHIVE IS OPEN, FREE, AND HOLDS THE REGISTERS. Its first response was a «Making sure you're not a bot!» interstitial that resolved by itself — precisely the behaviour reported. Ingouville, décès, 1843 returns five results including THE COMPLETE YEAR'S DEATHS REGISTER, COTE 4 E 9077, 310 VIEWS, with two tables décennales over it. Le Havre returns twelve, covering the town and the communes it absorbed. A French death act carries the age, the birthplace, the parents and the surviving spouse — MORE THAN THE CONSULAR RETURN THIS ARCHIVE WAS QUEUEING FOR, and for Margaret that is the nearest thing to a birth date for a daughter of Robert D'Arcy. Work list 141: read them, tables décennales first.
• AND BACK TO RESEARCH, BECAUSE THE DAY HAD FOUND FOUR OWN ERRORS AND NO ANCESTORS. The newly-opened source was asked a positive question, which is the one it can answer. HOLY TRINITY CHURCH, KINGSWOOD HOLDS THREE SANIGAR BURIALS — Andrew 1980, Dennis 2003, Barbara 2008 — and the cemetery page settles which Kingswood with a coordinate: 51.4616, -2.5033, which is KINGSWOOD IN BRISTOL, six kilometres east of the centre. Kingswood near Wotton-under-Edge lies twenty kilometres north at 51.63, -2.36. THE COUNTY COULD NOT HAVE DECIDED IT — both were Gloucestershire, the Bristol one until 1974 — and guessing from the name would have put this family's surname back in the Vale it came from rather than the city it moved to.
• THE CHURCHYARD IS ALSO A FINDING IN ITSELF: 571 records and 402 headstone images, free and without an account. It is the first Bristol burial ground this archive can read end to end. Twentieth-century, so it answers nothing about the eighteenth — what it shows is the surname still standing in Bristol in 2008, four generations after Thomas Sinegar was married there by banns.
• AND TOOK THE RECORD ATLAS SESSION'S CHALLENGE LITERALLY: does any probe here ask a different question from the one whose answer it records? Theirs had been a liveness check passing against a host that served nothing while 25 of 27 tests failed. Most of this archive's gates have been proved to fail in anger this fortnight — controls refused a `ctl` on a blocked row this week, atlas refused the Thornbury pin an hour ago, retired refused a restated reading, secrets refused a planted key, whosplit refused Miriam on its first run. THE ONE THAT HAS NEVER ANSWERED ANYTHING IS check_cites.py.
• IT IS WIRED INTO NOTHING AND HAS NEVER RUN. It tests the links OUT of this site, which nothing else does, and a dead citation looks exactly like a live one on the page. I wired it into build.sh and took it straight back out, because the measurement said not to: 575 DISTINCT EXTERNAL URLS, one at a time, 12-SECOND TIMEOUT EACH, AND NO OUTPUT UNTIL THE LAST RETURNS. A run was stopped at eight minutes having printed nothing. And 428 OF THE 575 ARE MYHERITAGE LINKS carried in from the GEDCOM export rather than citations this archive chose — as written it would spend most of its time checking somebody else's footnotes. Work list 140, with what it needs to become usable. A step people disable is worse than no step.
• AND THEN THE CORRECTION WAS WRONG TOO, WITHIN THE HOUR. It said SANIGER returns 25 burials and none in the United Kingdom. The same query at `size=50` returns FOUR IN ENGLAND: Andrew, Barbara and Dennis Sanigar all at HOLY TRINITY CHURCH, KINGSWOOD, and Maisie Sanigar at Scartho Road Cemetery, Grimsby. The «none» was the page size. Twentieth-century burials, so not evidence for the eighteenth-century questions — but three of this surname in one churchyard, and WHICH Kingswood is unresolved: there is one in Gloucestershire three miles from Wotton-under-Edge and another in Bristol.
• THE SITE CANNOT BE SEARCHED BY URL AT ALL. `cemetery_country=United+Kingdom` is accepted and ignored — identical rows with and without it, French and American and Australian among them, which also means the SMITH «control» published an hour ago was never filtered. `page` is accepted and ignored: `page=7` returns the same first three records as `page=1`, and the footer reads «Page 1 of 200» whatever is asked. About ten thousand D'Arcy records exist and the first fifty are all a URL can reach. THE FOURTH SEARCH PARAMETER THIS ARCHIVE HAS FOUND ACCEPTED AND IGNORED, after FindMyPast's `place`, `sourcecategory`, and `yearofbirth` with `keywordsplace`. Usable for a positive, useless for a negative. [Own error 54](/corrections/).
• AND A SOURCE THIS ARCHIVE HAD WRITTEN OFF IS OPEN. The Record Atlas session mentioned in passing that eight French departmental archives in its survey refuse a SCRIPTED request and open normally to a real browser, and suggested retesting anything written off that way. Four rows here carry «blocked». Two are account walls and genuinely shut. IRISHGENEALOGY.IE SURVIVED THE RETEST — still sitting on a Cloudflare interstitial after twenty-two seconds, not defeated and not going to be. AND BILLIONGRAVES WAS SIMPLY WRONG: it answers a browser at once, and the 403 and 400 behind the verdict were facts about the request.
• WHAT IT HOLDS, NOW IT CAN BE ASKED: SANIGER returns 25 burials — United States, Canada, France, Australia — AND NONE IN THE UNITED KINGDOM. Control: SMITH filtered to the United Kingdom returns Swarkestone and Portslade in England and Forfar and Aberdeen in Scotland, so British cemeteries are in the index and this is a null rather than a gap. A weak null: the site is volunteer-photographed and thin in Britain, so it can support «not found here» and never «not buried». [Own error 53](/corrections/).
• WENT BACK TO THE 24 PARENT LINKS LEFT UNDRAWN, and the first measurement of them had been against the wrong index. It counted «parent not in the tree» using ancestors.json, WHICH HOLDS ONLY DIRECT-LINE ANCESTORS — Joseph D'Arcy reads as «0 in tree» there while having a person page. Asked of the GEDCOM's 15,643 individuals instead, the picture changed, and seven more links were safe to draw.
• JOSEPH D'ARCY'S SECOND FAMILY IS NOW DOCUMENTED ON THE CHARTS. Catherine Lucy Jane, John Hyde, Josephine, Frank Hyde and Joseph William were each baptised at Milford to «JOSEPH, Lt Col RA», and every one of their charts drew that father as «tree, untested». Charlotte D'Arcy's Chatham baptism of 1826 and Jane D'Arcy's christening at St Michael, Barbados in 1788 — the latter naming Robert D'Arcy as her father — are drawn too. 54 recorded relationships across 51 people, from 41 across 35 this morning. The mother of the Milford five is KATHERINE LUCY ELIZA in the register and is in the tree under no such spelling, so only the father is drawn: naming her would be guessing which of the tree's women she is.
• AND A REFUSAL NOBODY HAD READ. The REL table carried «Charlotte Maria D'Arcy» pointing at George Pitt D'Arcy, and the tree calls her CHARLOTTE D'ARCY — so build_register.py had been printing «ambiguous or absent, so no edge was drawn» for her on every build, and the edge had simply never existed. It is drawn under the right name now and the dead entry is gone, because a refusal that fires every build and is never acted on is the «ok» line in a different costume.
• AND THE SAME GAP WAS ON THE OTHER TWO TREE-DERIVED PAGES, so it was fixed as a class rather than a third time by hand. /graves omitted 26 documented burials and /households 8 documented census placements. The graves list includes JOHN SANIGER, HANNAH'S FATHER, BURIED AT BERKELEY ST MARY THE VIRGIN ON 16 JANUARY 1825, with Thomas Saniger of 1819 and Edward Saniger of Saniger, Gent., of 1739 beside him. /graves was never dishonest about it — it says «gathered from the tree» in its own first sentence — but saying where a list comes from is not the same as showing what it leaves out.
• THE COMPUTATION LIVES IN ONE PLACE, site/src/lib/treegap.js, and /marriages was refactored onto it rather than left with its own copy. Three copies of one rule drift, and a surface quietly showing a different difference from its neighbour would be worse than the fault being fixed. All three now read «And N the tree does not carry» — 34, 26, 8 — computed each build, so documenting one more puts it on the page by itself.
• THEN THE CHARTS. 30 of the 58 record lines that name a parent belonged to people whose pedigree drew no edge at all — documented parentage still rendering as «tree, untested». The cause is structural: REL is hand-kept and the record lines are generated, so a new record never becomes an edge unless somebody writes one. Six were added and drawn, taking the charts from 41 edges to 47 across 44 people. ONLY DISTINCTIVE FULL NAMES WERE USED: the automatic pass also offered «Mary», «Elizabeth» and «MARY» resolving to exactly one tree person each, and this family reuses forenames without mercy. Two of the eight proposed were refused as ambiguous by the tool itself and are named rather than drawn.
• THEN THE ATLAS, WHICH CANNOT MAP ANY OF THE SEVEN PLACES THIS ARCHIVE HAS DOCUMENTED AND THE TREE DOES NOT HOLD. Four resolve to 52.531,-1.265, the geocode for the word ENGLAND, each stamped «exact»: Bourn (Catherine Georgiana West's burial), Lymington (Frank Hyde D'Arcy's marriage and Joseph D'Arcy's death district), Penzance (George Abbas died there), Milford (two D'Arcy baptisms). Two resolve to nothing: Bangor, where Robert West D'Arcy died, and Bombay, where he married.
• AND THE SEVENTH IS THE ONE WORTH RECORDING. Thornbury looked like the single good case and was added to the map with Lydia Sanigar's death registration behind it. check_atlas.py REFUSED THE BUILD: Thornbury and Gloucestershire now shared 51.833,-2.167, and neither contains the other. It is the county centroid standing in for a market town twenty miles away. OWN ERROR 41 ATTEMPTED TWICE IN ONE HOUR, CAUGHT THE SECOND TIME BY THE GATE WRITTEN AFTER THE FIRST — and caught after I had already congratulated myself on avoiding it. The entry was removed; all seven are counted on the work list instead of guessed at.
• THE TIMELINE IS HONEST AND WAS LEFT ALONE. Its dek says the point «is not when people were born — it is what was happening around them when they were», which is a stated curation rather than a claim of completeness, so a documented event missing from it is not a fault. /graves says «gathered from the tree» in its own first sentence. Those two needed nothing.
• THEN THE SAME QUESTION OF THE DIAGRAMS, and /marriages was overclaiming. Its dek read «EVERY MARRIAGE THIS ARCHIVE CAN DATE, in order». The page is built from marriages.json, which build_family_pages.py derives from the GEDCOM — so what it actually showed was every marriage THE TREE can date. The difference is 34 marriages this archive holds a record for and the tree does not carry, AND ONE OF THEM IS HANNAH SANIGER'S OWN MARRIAGE TO JAMES WAKEFIELD AT BRISTOL ON 17 OCTOBER 1831, which is the founding marriage of the English line.
• THE CLAIM IS CORRECTED AND THE DIFFERENCE IS NOW PUBLISHED. The page computes it at build time from person-records.json and dossiers.json against the couples table, so it cannot go stale: document a marriage and it appears there until the tree catches up. Among the 34 are William Sneyd and Mary Blackbourne at Madeley in 1769, Robert West D'Arcy at Bombay in 1846, Frank Hyde D'Arcy at Lymington in 1863, and the 1852 Bristol marriage of James Sanigar found a fortnight ago.
• AND THE FIRST DRAFT OF THAT BLOCK MADE ITS OWN FALSE CLAIM: it said «the first of them is the founding marriage of the English line» while sorting the table alphabetically, which put Agnis Sainger at the top. Caught by reading the built page rather than the intention. /graves and /households were checked in the same pass and neither overclaims — graves says «gathered from the tree» in its own first sentence.
• ASKED WHETHER THE RESEARCH REACHES THE PEOPLE AND NOT ONLY THE PAGES. Mostly yes, and four people were missed in two different ways.
• THREE HAD NO RECORD LINE AT ALL, so neither surface could show them: CATHERINE SNEYD, who died aged four — death registered December quarter 1849, Stoke on Trent, GRO volume 17 page 114, with a national control returning exactly one such death in eight years; JAMES SANIGAR, whose 1852 Bristol marriage this archive found on 27 September; and LAURA SAYER, who is on the same page of the index and is almost certainly his wife. All three were written up, graded and published in prose, and none of them existed as a person anywhere on the site. The evidence gate reads register.json and checks that every line reaches its person — IT CANNOT SEE A PERSON WHO WAS NEVER GIVEN A LINE.
• AND ONE WOMAN WAS SPLIT IN HALF. Joseph D'Arcy's first wife had two /who pages: «Lady Catherine Georgiana West» carried her two death notices, «Catherine Georgiana West» carried her burial and the Westminster baptisms of her sons. A reader landing on either saw half of her. The second line is this archive's own, added on 23 September under a name that already existed in another form.
• A NEW GATE, AND IT FOUND A SECOND ONE ON ITS FIRST RUN. check_whosplit.py asks whether two dossiers differ only by a title or a trailing qualifier — «Lady X» beside «X», «X, née Y» beside «X» — which is the shape this archive actually produces. It immediately refused on MIRIAM WAKEFIELD, who is IN THE TREE and whose two most important records were parked on /who pages because the register filed them under her married forms: her sworn deposition, and her 1909 death registration NAMING HER MOTHER AS HANNAH SANIGER, which is the link into the Saniger line. Her person page carried four records and now carries six.
• build_dossiers.py already guards against this and says so in its own comment — «duplicating them would split the same human in two» — but it compares the tree's names against the register's EXACTLY, so two different strings for one person never meet the test. The guard was real and the fault walked around it.

found Added/errands 35 — the two 1843 deaths at Havre have a free French route, never tried, and it is not KewAdded/sanigers — three Sanigar burials at Kingswood, and a coordinate settling which KingswoodCheckedevery gate, for one that answers a different question from the one it reports — and found a tool that asks nobodyCorrected/coverage — the BillionGraves correction was itself wrong. Four English Sanigars, not none. Own error 54Corrected/coverage — BillionGraves was never blocked. Own error 53Addedthe charts again — Joseph D'Arcy's five Milford children, Charlotte at Chatham, Jane at Barbados. 41 edges became 54Added/graves and /households now publish what the tree does not carry, as /marriages does — 26 and 8Addedthe charts — 6 more documented parent links drawn; 41 edges became 47Checkedthe atlas and the timeline — the timeline is honest, the atlas cannot map seven documented placesCorrected/marriages said «every marriage this archive can date» and showed only the tree's. 34 documented ones were missingFixed/people/miriam-wakefield — her death registration and her sworn deposition were on two other pagesFixed/who/catherine-georgiana-west — one woman, two dossiers, half her evidence on eachAdded/who — Catherine Sneyd, James Sanigar and Laura Sayer, documented for weeks and carried by nothing

Two deaths looked for, neither found, and the reason is worth more than the search

did

• CLEARED THE UNPUBLISHED BACKLOG, and the useful part is that «unpublished» turned out to be two different faults. Of eight passages the gate was reporting, TWO WERE THE CHECKER: sources.json carries literal markup that was stripped from the page side and not the data side, and the coverage legend's glosses live in title= tooltips, which tag-stripping discarded. That is the THIRD normalisation fault in this one tool, after the markdown and the whitespace.
• OF THE SIX THAT WERE REAL, FOUR WERE LOST AND TWO WERE DUPLICATES. The lost: /bristol-chapels never said what it was searching for, /behind-the-paywall and /saniger-line never named the record sets behind their findings, and /SANIGERS HAS CARRIED ITS SPELLING TABLE SINCE 21 SEPTEMBER WITH NO STATEMENT OF WHAT IT COUNTS — 27 spellings tried, 20 productive, 373 entries, one stated basis, and a floor rather than a total. All four published.
• THE TWO DUPLICATES WERE REMOVED, NOT RENDERED, and the near-miss is the lesson. portsea.theFind held Joseph D'Arcy's 1780 baptism as seven structured fields, and the page's own opening already says all of it, including «the only document anywhere that names Major-General Robert D'Arcy as somebody's father» — which is the same claim as the field I was about to publish. Rendering it would have printed that record twice on one page. The reason for each removal is written into the file it was removed from.
• WENT AFTER THE TWO SONS' DEATHS, because both died in Britain after civil registration and the GRO index is free through FreeBMD. Neither can be found there, and in both cases the quarter this archive needs is the one that is not transcribed.
• GEORGE ABBAS KOOLI D'ARCY died on 22 October 1885, which registers in the December quarter. DARCY deaths for 1885 return March and June and stop. Control: SMITH deaths in the Penzance district across 1885 return March, June and September, each marked >99% transcribed, and NO DECEMBER. The gap is the quarter, not the surname.
• ROBERT WEST D'ARCY died at Bangor on 14 June 1862. DARCY deaths for 1862 return March, September and December and no June — which alone could mean either no D'Arcy died that quarter or the quarter is missing. Control: SMITH deaths in the Bangor district across the WHOLE of 1862 return zero rows and not one quarter header, which cannot be a real result for a district that size. Bangor 1862 is simply not in the index.
• SO NEITHER IS A NULL, and the coverage row now says why: FreeBMD prints its completeness beside a quarter it has — «(>99%)» — and prints NOTHING AT ALL for a quarter it does not, which is the one case that looks like evidence. Before quoting an absence from it, run a common surname at the same district and quarter.
• A NOTE ON METHOD, because it nearly went wrong twice. The first run of the 1885 control was contaminated — a stray click had put SMITH into the spouse field as well, narrowing the result silently — and was re-run clean. And the search form's pixel coordinates could not be trusted: the screenshot frame changed height between calls, 505 then 600 then 500 against one viewport, so clicks computed from one frame missed in the next. The form was driven through its own controls instead, with every field read back before submitting.