Ontology Languages and Vocabularies Adopted

Context

OPDA’s linked-data work needs a declared, bounded set of ontology languages and vocabularies. Without a published list, modellers will reach for whatever they know — producing redundant terms, unprincipled mixing of W3C Recommendations with community drafts, and an unreviewable surface area.

The H&M semantic-modelling programme has spent two years pressure-testing this question across roughly 90 ODRs and 250 Council sessions. A survey of every @prefix declaration in its src/ ontology (~90 .ttl files) identifies 18 external standard vocabularies in active use. Every vocabulary admitted here is a W3C Recommendation, a maintained community standard, or a research-community ontology with broad linked-data uptake; the novelty is the OPDA-specific scoping — which vocabularies we adopt, with what conditions, in which layer.

The question this ODR answers: what closed, tiered catalogue should bound OPDA’s vocabulary surface, and on what adoption discipline?

Decision

Adopt a three-tier survey-grounded catalogue — Core / Conditional / Defer — porting the H&M src/ survey and rescoping for OPDA. It is the only option that bounds the vocabulary surface to a reviewable, authority-grounded set while documenting non-adoption as durably as adoption.

Decision rationale — options evaluated

The three options considered and why only the tiered catalogue was adopted.

Vocabulary tier structure

The three tiers and the direction of promotion/demotion that the catalogue governs.

Rules

Core — adopt unconditionally

The RDF stack and the small set of vocabularies every OPDA linked-data file is expected to use.

Conditional — adopt where the use case is present

Admitted only in the layers/files where the corresponding modelling concern arises. Outside those layers they are not used (SHACL gates will be added to enforce this — see H&M ADR-0147 R12 for the pattern).

Defer — reviewed and not adopted (yet)

Listed explicitly so future modellers know the question has been asked and the answer was “not now.”

Promotion and demotion criteria

Added by Session 002 Q3, drawing on the DCMI Usage Board admission test (Baker, Bechhofer, Isaac, Miles 2013), FIBO Production-tier discipline (Kendall+Davis), W3C TAG cool-URIs persistence (Hendler), and Cagle DA’s operational-check demand. 9-0 vote.

Conditional → Core promotion. ALL FOUR conditions must hold:

1. Named consumer. At least one OPDA module ODR cites the vocabulary in its ## Rules, with the vocabulary’s terms appearing in published Turtle (not just plan-stage prose).
2. Layer count. Used in ≥3 independent OPDA modules / layers.
3. SHACL gate. A SHACL gate enforcing the Conditional-layer scope has been published (per H&M ADR-0147 R12 pattern referenced in ### Enforcement).
4. Failure-mode test. A diagnostic exemplar (per Session 001 Q1 amendment lineage) where removing the vocabulary causes a specific named test to fail, demonstrating load-bearing work rather than decorative annotation.

Demotion is asymmetric (Allemang+Hendler joint position; Hendler’s preserved Scope-Check 1 Q4 audit-trail concern):

• Core never demotes. URI-graph break — every downstream module dereferences Core, every published header includes its prefix. Deprecation is recorded in ## Change log (the term is dcterms:isReplacedBy-style retired) but the row stays. W3C Process precedent (REC + ERRATA + REC-revision keeps the lineage visible).
• Conditional → Defer is editorial; requires (a) non-use across one full Phase OR (b) Core entry now provides the semantics. Named voter; Change Log row attribution mandatory.
• Defer rows never delete. Audit-trail discipline — reviewed-and-not-adopted is a governance act; row stays so future maintainers don’t re-litigate. FOAF / BBO / ArchiMate are canonical.
• One-step-per-Change-Log-row (Hendler sub-rule): tier movements cross one boundary at a time (Defer → Conditional → Core; no Defer → Core skip).

De-listing (Defer → out of catalogue) is reserved for:

• The W3C / maintainer formally withdraws the vocabulary.
• The OPDA WG (or adopting project’s governance per ODR-0001 §Adoption) rules the vocabulary out by name.

Otherwise Defer entries persist indefinitely — the recurring-question record is the value.

Review cadence (Davis position). Annual author-only review reads current W3C status, refreshes the W3C status field per entry, and proposes movements per the four-condition test; only contested rows escalate to Reduced Council. New-vocabulary admission requires Reduced Council minimum (per ODR-0001 §When to use the Council).

Profile-pinning ownership

Added by Session 002 Q5. Singapore Framework’s DCAP-by-consumer pattern (Nilsson, Baker, Johnston 2008) is the precedent. 9-0 vote with Cagle DA full withdrawal.

When an admitted Conditional entry points at a profile slice of a large upstream vocabulary (DPV’s dpv-pd slice; an eventual FIBO module slice; an ODRL Common Vocabulary slice; PROV-O qualified-attribution forms), the profile authoring rule chain is:

1. Module proposes. Profile-pin proposals originate in the consuming module’s Council session (e.g. ODR-0012 proposes dpv-pd slice; ODR-0010 proposes DASH-for-form-driving slice; ODR-0009 proposes PROV-O qualified-attribution slice).
2. Catalogue records. This catalogue records the pin in the entry’s Profile pin column with attribution to the consuming session, and adds a Change Log row.
3. Module veto. Module owners retain veto over pins affecting their shape graphs.
4. Cross-module conflict default to union. Where multiple modules consume the same vocabulary with different profile needs (e.g. DASH for ODR-0010 form-driving AND ODR-0013 identity-key validation), the catalogue records the union of pinned slices, not the most restrictive.
5. WG ratifies disputes only. The adopting project’s WG ratifies only when modules cannot agree; otherwise the chain is module-proposes → catalogue-records → consumed.

The catalogue does NOT author profile shape internally.

Reference-not-import (normative)

Added by Session 002 Q4. Hoists Adoption-pattern rule 3 to first-class normative status. 9-0 vote with Cagle DA full withdrawal on the three-value adoption-mode field qualification.

Every Conditional-tier entry adopts by reference, not import, as the default MUST. Each row declares its adoption-mode:

A row MUST justify any choice other than reference-only with a one-line rationale in the row’s Notes column. The Council session that authored the choice attributes via ## Change log.

Rationale: reference-only is the FIBO discipline (fibo-fnd-utl-av references dct: without owl:imports); the symmetric W3C TAG persistence rule applied to imports — don’t import URIs you haven’t committed to maintaining. Pandit’s ODR-0012-side concern on dpv-pd bundled-import for runtime PII hierarchy validation is recorded as an ODR-0012 implementation concern (the catalogue rule is reference-only; ODR-0012 may author slice-import when the lawful-basis class vocabulary surfaces).

Change log

This catalogue is governed in place: amendments to tiering or rationale are recorded as rows here, attributed to the Council session that authored them. The amendment-ODR pattern (formerly ODR-0014) is retired by Scope-Check 1 (2026-05-26, Q4 vote 7-1-1) on FIBO / DCMI / W3C-WD-discipline grounds; provenance is preserved here, not in a parallel record. Hendler’s dissent on the retirement (“every governance act stays permanently”) is recorded in the follow-up plan’s risks (docs/plan/council-followup-sessions.md §9), not silenced.

Adoption pattern (applies to every Conditional entry)

1. Canonical URI — use the vocabulary’s published namespace, not a local re-mint.
2. Local SHACL enforcement — write SHACL shapes in the relevant OPDA layer/file that constrain usage (cardinality, datatype, severity).
3. No owl:imports — reference by URI only; let external consumers fetch the upstream ontology themselves.
4. vann: header on every owl:Ontology — declare the preferred prefix so dereferencers can render snippets consistently.
5. Recorded provenance — when an external vocabulary appears in a new layer for the first time, the introducing commit or ODR cites the use case.

Enforcement

• Each row above declares canonical URI, role, and constraint; modellers MUST reference the canonical URI and MUST NOT re-mint terms in opda: that duplicate adopted-tier semantics.
• Conditional-tier gating is honour-system until SHACL gates are written; the follow-up is tracked in docs/governance/deferred-work.
• The Defer tier is reviewable on a schedule (annual, or whenever a triggering use case arises). Promotion/demotion is recorded in ### Change log above, attributed to the Council session that authored the change.
• The amendment-ODR pattern is retired: changes to this catalogue land as new rows in ### Change log, not in a parallel record.

Alternatives

• “Use whatever vocabulary fits the modeller’s preference” — produces redundant terms and unreviewable surface area within a year.
• “Adopt every vocabulary in active linked-data community use” — surface area becomes a multi-person-year mapping problem with no business return.
• “Reinvent the necessary terms under an opda: namespace” — discards two decades of W3C linked-data work and isolates OPDA outputs from external consumers.

Consequences

• Reference the published catalogue when introducing any external vocabulary; do not debate the choice per file.
• Use canonical, dereferenceable URIs throughout — external consumers depend on them resolving.
• When a recurring “why don’t we use schema.org / FOAF?” question is raised, point at the Defer column; do not relitigate without a triggering use case.
• Write SHACL gates for Conditional-tier scope as soon as the H&M ADR-0147 R12 pattern is portable; until then, conditional-tier compliance is honour-system and reviewers MUST check it manually.
• Pin versions explicitly in ODRs only where currently declared (RDF 1.2, SHACL 1.2). When DPV / DCAT 3 / DASH undergo a breaking version change, raise a follow-up ODR.
• Keep the catalogue alive: when a vocabulary good for OPDA but absent from H&M is proposed, do not reject on “no precedent” alone — amend the catalogue.
• Treat ArchiMate and BBO as candidates for demotion to Defer at the first Council review if no process- or capability-modelling use case has materialised.

ODR relationship graph

This record’s position in the ODR dependency chain, as declared in the frontmatter.

References

• Catalogue change log lives in ## Rules above. Sessions amending the catalogue: session-001 Q2 (multi-row amendment); Scope-Check 1 Q4 (governance-pattern: retire ODR-0014 — fold here) and Q7c (admit cred:, did:).
• Superseded artefact: ODR-0014 — formerly carried the Session 001 amendments as a partial-supersession record; retired 2026-05-26 per Scope-Check 1 Q4 (vote 7-1-1; Hendler dissent preserved). ODR-0014 retained as historical anchor for Council Session 001 provenance; its content is folded into ### Change log above.
• FOAF — ruled out. Session 001 Q2 briefly reopened the Defer-tier FOAF entry (because prov:Agent is deliberately thin — no person/organisation distinction, no structured name), but FOAF has since been ruled out. The Kind layer uses the W3C Org ontology or a bespoke opda: model, with prov:Agent for the provenance role only. Settled in ODR-0006; recorded in the change log above.
• W3C VC / DID Compatibility Layer: cred: and did: admitted to Defer per Scope-Check 1 Q7c; activation deferred to ODR-0016.
• Provenance: catalogue ported from a survey of the H&M src/ ontology @prefix declarations. The adoption pattern (canonical URIs + local SHACL + no owl:imports) is inherited from H&M ONT-0071c/i/j and ONT-0086.
• Related: Council methodology ODR-0001; programme anchor ODR-0003. Relates contextually to ADR-0001 (DCAM/DMBOK adoption); not a typed dependency.