Bulk downloads
/dump/ carries the whole registry, rebuilt on every deploy. Read it instead of crawling the record pages: one request replaces thousands. The entry point is datapackage.json, a Frictionless descriptor that names every file with its byte count and its sha256 hash. Read the sizes and the hashes there: this page lists the files, and the descriptor is their machine-readable record.
These artifacts describe registry version 0.1.0. The records are the same ones that /id/ serves, without the JSON-LD @context. The API contract documents the media types: the host serves a .jsonl body as application/octet-stream, so parse by documented shape, not by response header.
| File | Contents |
|---|---|
works.jsonl | One Work record per line: the texts that get cited. Read line by line. Each line is one JSON object. |
citation-systems.jsonl | One CitationSystem record per line: the locator grammar of a citation tradition. Read line by line. Each line is one JSON object. |
references.jsonl | One CanonicalReference record per line: a single point in a work under one citation system, with its resolver targets. Read line by line. Each line is one JSON object. |
mappings.jsonl | One MappingAssertion record per line: a link from a work to an identifier in another system. Read line by line. Each line is one JSON object. |
aliases.json | The whole alias table as one JSON object. A key is a /cite/ alias path or an external identifier IRI; a value is the canonical IRI it resolves to. Read as one JSON document. |
datapackage.json | The manifest for the five files above: the byte count and the sha256 hash of each, plus the registry version and the licence. Read as one JSON document. |
Verify a download
Recompute the hash, then compare it with the hash field that datapackage.json records for that file. The descriptor writes it as sha256: and the hex digest:
curl -sO https://textrefs.org/dump/references.jsonl
shasum -a 256 references.jsonl
curl -s https://textrefs.org/dump/datapackage.json | jq -r '.resources[] | select(.path == "references.jsonl") | .hash'The six files above are registry data, and carry the CC0 1.0 dedication that datapackage.json declares, as all TextRefs registry data does. This page is not one of them: the site code that renders it is licensed AGPL-3.0-or-later.