Named RDF 1.2 reifiers for mapping provenance

Context and Problem Statement

Mapping metadata needs an unambiguous, browsable statement identity. Maintaining an asserted SKOS statement and a separately edited three-part description creates two representations that can diverge. Blank-node metadata records also lack stable public identifiers. A canonical triple-term carrier reduces duplication, but only works if storage, validation, queries and exports support the same representation.

Considered Options

• Canonical RDF 1.2 triple terms with named reifiers and derived exchange formats.
• Canonical OWL annotation-axiom triples.
• Defer the change until the selected standards and implementations mature further.

Decision Outcome

Use named RDF 1.2 reifiers in canonical storage; retain the asserted SKOS statement. Generate OWL annotation-axiom and SSSOM TSV exports from that storage. This decision governs the carrier; ODR-0056 governs SSSOM fields and cardinalities, including its exclusion of mapping authorship.

Consequences

• Good, because one statement-identity arc replaces independently maintained endpoints.
• Good, because a named reifier supports stable links and evidence inspection.
• Bad, because RDF 1.2 storage and SPARQL accessors need explicit runtime evidence.
• Bad, because a migration must coordinate data, queries, shapes and exports.
• Neutral, because this improves internal consistency without claiming a new external consumer capability or widening the older candidate’s RDF Basic conformance claim.

Confirmation

Implementation is pending in OPDA. All B1–B7 conditions below must be demonstrated on the actual target runtime. Historical measurements on another system do not prove local parser, validator or query support. A successful empty query is a failure when populated provenance is expected.

Rules

Canonical pattern

An asserted SKOS mapping and its named sssom:Mapping reifier coexist. The reifier has exactly one rdf:reifies triple term identifying that same subject, predicate and object. Quoting/reifying a statement does not assert it. Validate equality between the identified statement and the retained assertion; the representation does not eliminate the need for that check.

B1: Stable named reifiers

Use stable full IRIs, not blank nodes. The named IRI is authoritative. Within a mapping set, the collision key is the exact subject-IRI, SKOS-predicate, object-IRI tuple. A different predicate represents a different assertion. Curator identity is not part of this identity key. Any identifier minting must avoid local-name collisions and respect Turtle syntax.

B2: Preserve the asserted layer

The corresponding SKOS assertion MUST remain present in every committed migration state. Provenance consumers must not be served by removing predicate consumers’ data. Retain the assertion even when a triple term identifies it elsewhere.

B3: Enforced predicate authority

Use a sh:targetClass sssom:Mapping node shape for metadata and an executable SPARQL constraint over rdf:reifies for triple-term and predicate checks. The selected approach uses IS_TRIPLE() and PREDICATE() accessors. A non-triple value or a non-SKOS-mapping predicate fails. An owl:annotatedProperty restriction checks an export, not the canonical store. Do not substitute it for the canonical gate.

sh:reifierShape is not selected for this profile: the source decision found it silently inert on its tested engine. That historical finding is a warning, not a claim about every current engine. Reopening requires a negative probe that actually fails on the target runtime and an explicit profile amendment.

B4: Self-falsifying validation evidence

Retain executable fixtures proving all three outcomes:

• A missing required SSSOM property produces nonconformance and a violation entry.
• A reified non-SKOS predicate produces nonconformance and a violation entry.
• A complete valid reifier conforms.

A parsed shape or a green run without demonstrated rejection is insufficient. Separate severity contracts when the runtime does not honour per-constraint severity as expected; do not silently downgrade required metadata failures.

B5: Populated query results

Every governed mapping must be returned with its required provenance by the consumer query. Test populated source, versions, release, justification and date, not just row count. Do not require the excluded sssom:author_id. Preserve named graph scope and do not suppress legitimate asserted mappings by a lexical-order filter intended to deduplicate symmetric relations.

B6: Atomic migration

Change canonical records, shapes and consumer joins together in one coherent commit. No intermediate store/query mismatch is acceptable. A compatible temporary reader, if needed for deployment, is explicitly bounded and retired after migration; it must not become a second source of truth.

B7: Regenerable exports

Generate the OWL owl:annotatedSource / owl:annotatedProperty / owl:annotatedTarget representation and .sssom.tsv from canonical Turtle. These are derived artefacts, never independently edited inputs. Delivery requires a CI round trip from canonical mappings through TSV back to the same mapping set, plus equivalence checks for the OWL view and permitted SSSOM fields.

Analytical and implementation boundaries

The source method analyses the objectified mapping relationship as a dependent relator and realises its record as an RDF reifier. This analytical description does not import a foundational ontology or turn the metadata record into either endpoint. Its technical identity is the named IRI and scoped tuple contract above.

The upstream deployment-risk objection remains relevant: compatible exchange formats do not erase the cost of changing canonical storage. Review actual standards maturity, parser/accessor support, validator rejection evidence and the functional benefit before changing this carrier. A future standards milestone alone does not prove that the chosen engine executes the required validation.

More Information

• SSSOM profile
• Source-method record: ODR-0096, revision 67174057e6384b79d0b28b7736fe70a66e112895.

Amendments

• 2026-09-05 — OPDA adaptation. Preserves B1–B7, the canonical/export split, validator-executability guard and live deployment-risk concern. Replaces source migration counts and execution receipts with local obligations; applies the later SSSOM authorship exclusion without changing the carrier.