gnomonlearn
Docs/Project manifest RDF identity proposal

Project manifest RDF identity proposal

Versioned draft snapshot

RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.

Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.

This is a trace and gap record from the pinned draft specification. It does not establish proof or implementation conformance.

Status: owner approved a separate manifest IRI and SourceFile class for all 20 file nodes; source projections and generated Turtle corrected. No public RDF release or deployed consumer requires a data migration; exact-IRI and exact-class clients remain possible and must be checked before publication.

RFC-0008 defines project as a DeclarationKind with FQN project and the subject https://gnomon.sh/domain/declaration/project. RFC-0017 uses that same subject for the project.gnm manifest, types it NamedDeclaration, and assigns FQN project.gnm. The ontology defines these as distinct sibling classes. Combining the RDF files makes one subject denote two source concepts with two canonical names.

The first correction retained the RFC-0008 subject for the declaration kind. Give RFC-0017’s manifest the subject https://gnomon.sh/domain/declaration/file/project.gnm, following RFC-0001’s existing /declaration/file/*.gnm family. It initially kept the NamedDeclaration class, project.gnm FQN, RFC-0017 provenance, and summary. No RFC-0017 source claim uses the old subject; its one domain node is the only RFC-0017 source occurrence. The source claim references to the old subject are in RFC-0008 and refer to the declaration kind, so they stay unchanged. The domain-link inventory mentions the old subject as a candidate for an unrelated RFC-0044 operation; that candidate remains valid for the declaration kind.

The implemented correction moves four RFC-0017 triples (rdf:type, definedIn, fqn, and summary) from the old subject to the new one. The old RFC-0008 declaration-kind triples stay at the old IRI. An in-memory counterfactual against the current full graph changed none of the 16 checked-in claim-query result lists (550 rows before and after). The source change regenerated only RFC-0017 domain Turtle. Projection drift, source-bound review, RDF-link, and claim-query gates pass. The domain identity gate now reports 196 shared IRIs, no rdf:type conflicts, 18 FQN conflicts, and 23 kindClass conflicts.

The later SourceFile correction resolves the file-class mismatch while preserving the separate IRI. Static search found the old IRI only in spec source/generated files, its framework duplicate, and spec review/inventory records; no executable source in the other extracted repos uses it. Exact-IRI external RDF consumers remain unverified. Nothing is deployed to migrate.

File-class follow-up

The RFC-0017 node is one of 20 domain/declaration/file/*.gnm nodes in the source-domain projection. RFC-0001 owns the other 19, and all 20 are typed NamedDeclaration. The ontology defines that class as a specific declaration instance and gives hasField a NamedDeclaration domain. A file can contain declarations without itself being one, so correcting only project.gnm would leave the same modeling error across the file family. The checked-in query, lint, Java KG, and projection tool sources contain no direct NamedDeclaration or declaration/file/ reference beyond a synthetic identity-gate test; this is a static local consumer observation, not an external RDF consumer audit.

The approved correction adds SourceFile under DomainEntity and retypes all 20 file nodes. Their IRIs, names, summaries, and RFC provenance are preserved; none asserts hasField. Source-bound reviews were refreshed, including the RFC-0017 match. The RDF diff removes exactly 20 NamedDeclaration type triples and adds 20 SourceFile type triples plus four ontology triples. The 16 checked-in claim queries run and do not reference the changed class or file subjects. Projection drift and review gates pass. The domain identity gate still reports no class conflicts, with 18 FQN and 23 kindClass conflicts elsewhere. Exact-class consumers outside the local corpus remain unverified. Nothing is deployed, so no stored-data migration is needed.