US Federal Acquisition Regulation Archive (Point-in-Time Editions)

United States · atualizações: weekly · us-acquisition-archive

Archived point-in-time editions of the Federal Acquisition Regulation and its agency supplements, collected from GSA’s own acquisition.gov archive — no API key, no login. One record is one regulation section inside one edition, keyed on the regulation, the section number and the edition’s effective date, so the same clause across successive editions is a series of records a buyer can diff. That key carries the citation exactly as a solicitation prints it — 52.212-4, 252.204-7012 — and rides in every chunk, which is how a procurement corpus joins to the clause text that was in force on its own date. The corpus deliberately includes the layer that is never codified in 48 CFR at all: the DFARS Procedures, Guidance and Information, the GSAM’s internal procedures and the other supplements’ non-regulatory halves. The currently effective codified text is not sold here — that is the US federal regulations dataset — because what this one adds is the time axis and the uncodified layer, not a second copy of title 48. Each record states the publisher’s own archive type, FAC number and effective date, the section’s heading, the part it sits in, whether its body is a solicitation clause, and the sections it cites. Editions are read from the publisher’s DITA packages, the only form that carries one file per section; editions published as HTML archives or PDF only are not included, and a section the supplement leaves empty is counted and skipped rather than shipped as a title with no text. Reproduction rests on 17 U.S.C. §105 and the edicts-of-government doctrine, and every record carries the source acknowledgement and the statement that an archived edition is not the currently effective text. Every record carries its source URL, fetch timestamp, HTTP status and the SHA-256 of both the raw response and the extracted record. Every release ships a manifest with every file’s SHA-256 and a Merkle root over every chunk hash, which the buyer can recompute from the downloaded files. record_id and chunk_id are stable across releases. Personal data is screened out inside the pipeline before any hash is computed.