← Documentation home

Canonical Markdown source · Oct 20, 2018

Canonical Identifier Policy

linked-art/canonical-identifier-policy.md · 88 lines · SHA-256 56bc7bc9ee2f

Status: active public-data policy

Last updated: 2026-07-05

Owner: Meta Museum data stewardship

Scope

This policy covers public Meta Museum Linked Art record identifiers, dataset

exports, dereferenceable API URLs, and persistent resolver strategy. It exists

so partners can reuse records without guessing whether identifiers are stable,

what redirects mean, or when a static dataset release should receive a stronger

external identifier.

Canonical URI Pattern

Canonical Meta Museum identifiers use:


https://lod.metamuseum.org/{type}/{ulid}

Source-provider URLs, Wikidata QIDs, ARKs, and other upstream identifiers are

modeled as equivalents, source ids, or identifiers. They are not silently

substituted for the Meta Museum canonical id.

Persistence Promise

Stable canonical ids are intended to persist for the life of the public dataset.

Identifiers are never recycled for unrelated objects. If a record is superseded,

merged, withdrawn, or removed because a source provider returns a legitimate

404/410 tombstone, Meta Museum should preserve an explicit redirect,

replacement, or tombstone trail rather than silently deleting the meaning of the

identifier.

Dereferencing And Redirects

Public record URLs should resolve to the current canonical JSON-LD record when

the record is active. If a record has moved, HTTP redirects should point to the

replacement. If a record is legitimately deleted upstream, the ActivityStreams

Delete/Tombstone evidence path should be used and the public surface should make

the removal state explicit.

Content Negotiation

Public record APIs support JSON and Linked Art JSON-LD reads. HTML pages expose

one canonical `describedby` link to their JSON-LD endpoint. Partners should use

`Accept: application/ld+json` when they need Linked Art JSON-LD and should treat

HAL links as navigation metadata, not graph assertions.

Dataset Exports

Machine-readable dataset metadata is published at `/api/datasets`. Versioned

record exports are available at:

  • `/api/datasets/exports/records?format=jsonld`
  • `/api/datasets/exports/records?format=ndjson`

Each export version includes a checksum. Each exported row preserves the Linked

Art record body and adds a packaging envelope for record-level rights and source

provenance. Exported records project object-level `equivalent[]` links from

stored equivalents, source provider URLs, provider public object pages, and

Wikidata entity URLs so consumers can reconcile through the standard Linked Art

path instead of digging into `_source` passthrough fields. Missing rights remain

review-required and non-reuse by default.

The dataset-level CC BY 4.0 license governs Meta Museum's export packaging,

metadata envelope, and dataset descriptor. Individual object/content/media reuse

is still governed by each row's record-level rights and source provenance; a

per-record CC0 right does not turn the entire export package into CC0, and the

dataset package license does not override stricter object/media rights.

ARK And DOI Strategy

ARKs are appropriate for object-level public resolver needs where an opaque,

long-lived identifier helps partners cite a record independently of source

provider URLs. DOI registration is reserved for curated static dataset releases

with fixed scope, declared license, release notes, and checksum, not for every

rolling API snapshot.

Partner Evidence Boundary

Identifier stability and export availability are local Meta Museum commitments.

External adoption evidence is counted only when a real partner reads or verifies

the relevant public surface. MetaHistoryBook has confirmed real Update coverage;

Delete coverage still waits for a legitimate upstream 404/410 tombstone and must

not be fabricated.