Why it exists

Almost everything needed to write this island's history already exists. It is held by university libraries, by federal and state agencies, by a historical society, by parish and township offices, by newspapers, and in private hands. Those custodians have done the expensive and unglamorous part — keeping paper dry for a century, paying for microfilm, running the scanner. Nothing here would be possible without them.

What does not exist is a place where all of it can be asked one question at once. A researcher who wants to know what happened on a particular stretch of shore, or to a particular family, or in a particular decade, has to know in advance which of a dozen custodians to approach, in what order, and in what format each will answer. The material is not missing. It isunassembled.

One place, and a copy that stays

The first job is custody: take a faithful copy of the document itself, record exactly where it came from, and keep it beside everything else. The collection now stands at 627 works — 18,709 files, 79 GB — spanning newspapers, books, manuscripts, charts and maps, government records, geospatial layers, sound recordings and archived web pages.

That matters because links die. Agencies reorganise their websites, hosts lapse, portals are replaced, and a citation that resolved last year returns a 404 this year with nothing to say where the document went. A URL is a pointer, not a copy, and an archive built on pointers inherits every outage of every host it points at. Every file here is stored with a provenance record naming its source, the date it was retrieved and a checksum of what arrived, so the copy can be shown to be the thing that was published even after the place it was published from has changed.

Reading the scans again

A great deal of this material was digitised ten or twenty years ago, when that was a considerable undertaking and the available tools were what they were. Those scans remain the primary artefact and are not improved on here. What has changed is everything downstream of them: text can now be recovered from a page image far more reliably than when it was made.

So held documents are read again with current optical character recognition and put through quality control before anything is published from them — the transcription checked against the page image rather than merely scored for plausibility, because a model that cannot read a damaged line will sometimes produce fluent, period-plausible prose instead of failing visibly, and that failure looks better on any statistical measure than honest, garbled output. 520 documents are also rendered page by page — 18,535 page images — so a reader can turn to the page a claim rests on rather than take a transcription's word for it.

Written, spoken, pictured, mapped

An island's record is not only paper. Alongside the documents this archive holds 6,373 media objects — 5,869 photographs, 186 nautical charts, 90 page extracts, 78 maps, 75 plates, 70 sound recordings and 5 moving images — and treats them as evidence of the same standing as a text. A photograph of a dock, a chart of the approach to it, a surveyor's field note describing the shoreline and a recording made a few hundred yards away are four accounts of one place, and they only work together if they are held together.

Geography is the spine that lets them join. 3,089 records carry coordinates or a drawn geometry, which turns a map from an illustration into something that can be queried: what else is recorded within this bay, along this road, on this island. 16,669 carry a date, so the same question can be asked of a decade. Between them, a place and a period are enough to gather material that arrived from a dozen different custodians in a dozen different formats.

Writing the relations down

Assembly is not the same as accumulation. A folder of files is still a folder of files. What makes the collection answerable is that each thing in it is given a shape — a person, a vessel, a road, a structure, an organisation, a chart, an event — with fields defined for that shape, and that the connections between them are recorded explicitly rather than left implicit in prose.

There are now 21,546 records across 45 shapes, joined by 64,808 stated relations: this person participated in that event, this event took place at that harbour, this chart depicts that reef, this vessel was operated by that company. Those relations are the difference between a search that returns documents and a question that returns an answer — and because every one of them is attached to a record that cites its own sources, a reader can follow any link back to the paper it came from.

Additive, not a replacement

Nothing here is meant to stand in for the institutions that hold the originals. Every item carries its provenance and points back at its custodian, rights are recorded as the holder states them, and material that cannot be published is described rather than shown. The aim is narrower and, we think, more useful: that the work of assembling this record only has to be done once, and that what results stays readable and stays put.

What is here sets out the holdings in detail; how it works describes the retrieval and provenance machinery; and getting at it is the reader's route in.