The Boston Archive
The record of one city, held whole.
The Boston Archive is the authority source for one subject. It holds the public civic record of Boston, Massachusetts — its places, people, organizations, events, works, and objects; the documents and datasets that attest them; and the grey record of what people say about the city that no authority ever wrote down. Every record carries a stable identifier, a canonical address, a full provenance chain, and its version history. Every record resolves as a page for a person and as JSON for a machine.
What this archive holds
How to read it — as a person
Every record has one address and one page. A place is at /place/{name}, a dataset at /dataset/{name}, a claim at /claim/{name}. Each page carries the record, the attestations made about it, its whole provenance chain, and every transform between the bytes as retrieved and the record as served.
Two things are worth knowing before you read. First, geography is held on two layers: the official boundaries an authority drew, and the lived names and extents people actually use. Neither overwrites the other; both are served. Second, the archive does not adjudicate. A claim records who attests it, since when, and on what evidence. Where attestations conflict, all of them stand, grouped and unranked, and you decide.
How to read it — as a machine
Append .json to any record address for the same record as JSON: /place/back-bay and /place/back-bay.json are one record in two renderings, built from one value. Every page also carries a schema.org graph under the CSTS-01 archive profile, with the source, the licence, the retrieval date, and the depth of the transform ladder addressable as graph nodes.
- /holdings — the holdings statement, and as JSON
- /ask?q=… — the retrieval surface, and as JSON
- /sitemap-index.xml — every record, addressable
- /feed.xml — accessions, as they land
/urn/{identifier}— resolve a citation back to its address
What may enter
Public-domain and open-licensed material only, with the licence recorded on every single record. Closed-licensed material does not enter, and the refusal is kept on the record beside the holdings so the collection can be audited for what it excluded as well as for what it holds.
AR-1. The retrieval surface is extractive: it returns records, ranked, with the passages that bear on your question and the citations behind them. It composes no prose. What you read at /ask was written by a source, not by a model.