Skip to content
VERASPEC
Repository
The VER Registrystable

The four value spaces: what is open, and what is only seeded

namespacevalue spacestatuswhat opening waits on
provenance_actionprovenance.chain[].actionopenNothing further. Draft §2.3 opens it; ratification of §2.3 is what remains
perceptual_algidentity.perceptual[].algseeded-closedThe revision that publishes the registry (draft §2.4)
source_classembeddings[].source.classseeded-closedThat same revision, and a registry-backed grade table (draft §2.4, E19)
segment_familymetadata.raw[].segmentseeded-closedThe revision that publishes the registry (draft §2.4)

Seeding is not opening, and the distinction is the whole point of the middle three rows. provenance_action is the only value space the 1.1 draft opens. For the other three the schema keeps its enum closed, and an entry in this snapshot is a description of the closed set, not an admission rule: a new perceptual algorithm or segment family still requires a schema revision, and registering one here would not make a Record carrying it valid. Draft §2.4 states the intent to open alg and segment in the same revision that publishes the registry; that revision has not shipped.

source_class carries one condition beyond the other two. Draft §2.4: source.class "additionally requires a grade assignment for each new class (E19) and MUST NOT be opened before the grade table is registry-backed." Snapshot 1 carries the grade table — every source-class entry states its grade — which satisfies the carriage half of that condition and activates nothing. Opening it is an act of the revision that opens it.

Two details of those seeded lists are load-bearing and easy to lose:

  • Grade tokens are ASCII, and caption_auto's grade is A- with HYPHEN-MINUS (U+002D). The typographic MINUS SIGN (U+2212) typeset in 1.0 §8's table is an artifact and MUST NOT appear in any interface, query parameter, index payload, or API response (E19). That codepoint appears nowhere in this directory, and tests/test_record_registry.py asserts it.
  • iptc and iptc-iim are one family. The metadata.availability key is iptc, the raw-segment family is iptc-iim, and a Consumer MUST treat them as equivalent (E21). The split is not repaired in 1.1 (draft §10.3) — carrying iptc as a deprecated alias would put two spellings of one family in VER 1.x for its lifetime — and the rename is deferred to VER 2.0. Each segment entry records its availability key, or null where the release set carries none.