Ontology namespace at w3id.org/opda/ via W3C PICG redirect

Scope clarification, 2026-09-07. This historical record governs the preserved schema-derived /pdtf/ identifiers. The new governed model instead follows ODR-0065: https://opda.org.uk/ns/ and https://opda.org.uk/id/. Neither the preserved corpus nor the isolated Property Pack candidate is reminted by that policy adoption.

✅ CURRENT IDENTIFIER SCHEME (definitive — amended 2026-08-22)

This block is authoritative; it supersedes every amendment, council placement, and the body below wherever they differ. Base https://opda.org.uk/; the stable identifier family for the schema-derived ontology lives under the historically minted /pdtf/ path with role/kind sub-segments. That path preserves RDF identity and dereferenceability; it does not name or imply an OPDA-endorsed PDTF standard or scheme. Slash, no hash, no version-in-IRI (technical ontology-release information remains in owl:versionInfo + owl:versionIRI → release snapshot). @prefix opda: <https://opda.org.uk/pdtf/> .

Resource	IRI
Ontology IRI	https://opda.org.uk/pdtf/ (+ owl:versionInfo "1.0.0")
Class / property	…/pdtf/Property, …/pdtf/evidenceType
SKOS scheme / concept	…/pdtf/scheme/role · …/pdtf/scheme/role/Buyer
SHACL shape node	…/pdtf/shape/EvidenceFacetShape, …/pdtf/shape/Baspi5_PropertyShape
Profile (form overlay)	…/pdtf/shape/profiles/baspi5
Named graph	…/pdtf/graph/foundation, …/pdtf/graph/inferred/entailment
data-dictionary entry	…/pdtf/harness/data-dictionary/propertyPack.environmentalIssues.flooding
ODR anchor	…/pdtf/harness/odr/ODR-0011/section-5a
ADR link	…/pdtf/harness/adr/ADR-0007-ontology-generator-specification
instance / test data	…/pdtf/harness/data/…, …/pdtf/harness/data/exemplar/<stem>
release snapshot	…/pdtf/harness/release/1.0.0/

Directing-authority overrides of the council (greenfield): everything nests under the /pdtf/ identifier root; SHACL shapes get a /pdtf/shape/ segment (overrides Q7’s no-/shacl/; this is an organisational sub-namespace of the schema-derived ontology, not a separate standard); named graphs get /pdtf/graph/; SKOS schemes + concepts get a /pdtf/scheme/ segment (ruling 2026-06-02 — disambiguates the scheme namespace from the flat term namespace, so a property opda:role at …/pdtf/role cannot collide with the role scheme at …/pdtf/scheme/role; applies to all schemes, not just colliding ones); /pdtf/harness/ nests (was a sibling). Profiles remain at /pdtf/shape/profiles/ as conformance shapes in that technical corpus; this placement does not make them normative SPDTF or evidence of OPDA endorsement. Retained from the council: slash, flat term namespace (no domain-module segments), no version segment, the technical-entity-vs-physical distinction (now core+scheme+shape+graph vs harness), Baker’s placement procedure, one-directional dependency (nothing in core /pdtf/ depends on /pdtf/harness/).

As-built (2026-06-02): the migration is implemented and green — opda_gen.namespaces is the single source of truth (OPDA, OPDA_SCHEME, OPDA_SHAPE, OPDA_GRAPH, OPDA_HARNESS + odr_ref/adr_ref/dd_entry/release_iri helpers + a flatten-collision guard). Per-module ontology IRIs collapsed to …/pdtf/graph/<module> importing the one ontology …/pdtf/; per-module/per-profile owl:versionIRI → …/pdtf/harness/release/…. dct:source → harness ODR/ADR/dd retained as provenance comments (not core→harness dependencies; the one definitional rdfs:isDefinedBy→ODR on opda:consumesFrom repointed to the core ontology). owl:Class-typed *Scheme (e.g. SpecialCategoryScheme, BoundedContextScheme) stay in the term namespace; only skos:ConceptScheme instances move to /pdtf/scheme/. Verified: 345 pytest + 8 CI gates (incl. byte-identity) + 27 repo-root round-trip. See docs/PLAN-2026-06-02-namespace-migration.md.

⚠ Amendment (2026-06-02) — base domain → https://opda.org.uk/ (supersedes the w3id/PICG choice)

On directing-authority instruction the base moves from the W3C-PICG path https://w3id.org/opda/ to opda’s own domain https://opda.org.uk/. This reverses this ADR’s core decision (and S004’s choice of w3id over an institutional domain for PICG persistence — the old Option A/C). Admissible on the same greenfield/pre-publication grounds as the 2026-06-01 amendment; the directing authority owns the persistence-vs-control trade for this build.

Mechanical effect: everywhere below (body + the 2026-06-01 amendment) read https://w3id.org/opda/ as https://opda.org.uk/ — the opda org path-segment is absorbed into the domain. The org/standard split is preserved as domain = organisation, path = standard:

• standard entities → https://opda.org.uk/pdtf/ (prefix @prefix opda: <https://opda.org.uk/pdtf/> .)
• physical resources + governance → https://opda.org.uk/harness/
• future second standard → https://opda.org.uk/<other>/

Infrastructure: the W3C PICG redirect (the original subject of this ADR) is no longer needed — opda controls opda.org.uk DNS + hosting directly and serves RDF from it. Bearing on session-037 Q2: opda-controlled hosting means per-term resolution / content-negotiation is fully in opda’s hands (the w3id flat-302 limitation that was the last objection to slash is gone) — the council’s slash verdict is reinforced; the hash recommendation’s resolution premise no longer applies.

(The redirect-target persistence trade — losing the PICG guarantee that survives org changes — is the cost; opda accepts it. Re-open trigger: if opda’s organisational continuity becomes a concern, reconsider a PICG redirect layer in front of opda.org.uk.)

⚠ Amendment (2026-06-01) — opda/pdtf slash-based scheme

On directing-authority instruction the namespace scheme is revised. This supersedes the hash commitment and the openpropdata.org.uk hosting/citation references in the body below (retained for provenance). The override is admissible because the ontology is pre-publication — no ecosystem-propagation cost yet applies (cf. the DPV slash→hash six-month precedent that motivated the original hash choice) — and the directing authority ratifies modelling decisions directly for this greenfield build. Where the body conflicts with this amendment, the amendment governs. Cross-corpus: ODR-0004 §Rules.1’s “single opda: HASH namespace” is correspondingly overridden and must be amended to match.

The scheme — four rules

1. opda is the organisation; pdtf is the standard. The ontology (Property Data Trust Framework) lives under https://w3id.org/opda/pdtf/.
2. Slash, not hash. Terms are per-term URIs; no # anywhere in the scheme.
3. No version segment in any IRI. Versioning is carried by an owl:versionInfo literal on the ontology, not the URL.
4. No module segment in term IRIs. One flat term namespace; the per-module TTL files are organisational splits, not URL segments.

Prefix: @prefix opda: <https://w3id.org/opda/pdtf/> . — so opda:Property → https://w3id.org/opda/pdtf/Property.

A second namespace, https://w3id.org/opda/harness/, replaces openpropdata.org.uk. The dividing line is standard entities vs physical resources:

• …/pdtf/ holds only the abstract published standard entities — the ontology classes and properties, and the SKOS concept schemes + concepts the standard defines. These are what a consumer cites as the PDTF standard.
• …/harness/ holds everything else: physical resources (the data-dictionary’s concrete PDTF data-field entries; instance/test data; named graphs) and the governance/build apparatus (ADRs, ODRs, exemplars). The data-dictionary is not a standard entity — its entries are the physical PDTF data fields, so they live in the harness, not /pdtf/.

Canonical mapping (old → new)

Kind	Old	New
Ontology term	…/opda/#Property	…/opda/pdtf/Property
SKOS concept	…/opda/#role/Buyer	…/opda/pdtf/role/Buyer
Concept scheme	…/opda/vocabularies/role	…/opda/pdtf/role
data-dictionary entry (physical PDTF data field — not a standard entity)	…/opda/data-dictionary#propertyPack.x.y	…/opda/harness/data-dictionary/propertyPack.x.y
Version IRI	…/opda/1.0.0/	removed → …/opda/pdtf/ + owl:versionInfo "1.0.0"
Per-module ontology IRIs	…/opda/property, …/opda/property-shapes, …/opda/property/1.0.0/	single flat …/opda/pdtf/ namespace (no module segment)
Profile	…/opda/profiles/baspi5/0.1.0/	…/opda/pdtf/profiles/baspi5
Named graph	…/opda/graph/foundation	…/opda/harness/graph/foundation
Instance/test data	…/opda/data/baspi5-conformant/…	…/opda/harness/data/baspi5-conformant/…
ADR citation	openpropdata.org.uk/adr/ADR-NNNN-slug	…/opda/harness/adr/ADR-NNNN-slug
ODR anchor	…/opda/odr/ODR-NNNN#section-5a	…/opda/harness/odr/ODR-NNNN/section-5a
Exemplar anchor	openpropdata.org.uk/data/exemplar/<stem>	…/opda/harness/data/exemplar/<stem>
SHACL shape node (base + profile)	…/opda/#PropertyShape	…/opda/pdtf/PropertyShape (co-normative; no /shacl/) — session-037 Q7
SHACL shape / profile document	…/opda/property-shapes, …/opda/profiles/baspi5 (doc IRI)	…/opda/harness/… (graph/document) — session-037 Q7/Q8
Profile (form overlay)	…/opda/profiles/baspi5/0.1.0/	default …/opda/harness/profiles/baspi5; …/opda/pdtf/profiles/baspi5 only on warranted adoption — session-037 Q8, operator call

Implementation note — multi-file ontology identity

With no module segment, the six module TTLs can no longer each carry a distinct …/opda/<module> owl:Ontology IRI. Resolution: the published ontology is one owl:Ontology at https://w3id.org/opda/pdtf/ (declared once); the module files contribute terms to it and are not separately addressable in the term URL scheme. Any document-level identity the build needs (e.g. per-file provenance) uses the harness namespace, never the /pdtf/ term space.

Redirect infrastructure (revised)

One W3C PICG redirect on the owner directory — /opda/ → the hosting target — with /opda/pdtf/… (standard entities) and /opda/harness/… (physical resources + governance/build) as paths the host resolves (council session-037 Q5: the harness is co-stewarded with pdtf, not separately delegable, so it is a path not a second coordinate namespace). Two coordinate redirects only if hosting genuinely requires pdtf and harness on different targets — a deploy-side detail this ADR already decouples from URI policy. The original single /opda/ → openpropdata.org.uk/ontology/ rule below is superseded. (If the directing authority accepts the council’s Q2 hash recommendation, the host serves the single pdtf document and per-term hash fragments resolve client-side — no per-term resolution build needed.)

Council session-037 dispositions (2026-06-01) — proposed, operator ratifies

Session 037 (Full Council; Queen Kendall, DA Davis; Baker, Gandon, Allemang, Knublauch, Pandit) reviewed this scheme against best practice + the hm and DPV prior art. Decisive criterion (Baker, all-voices): governance/identity boundaries in the URI; logical/serialisation/version distinctions in the triples — never in the URI string. The base path (Q1, 7-0-0), version-out-of-IRI (Q3), flat namespace (Q4, 6-1-0), and the standard/physical distinction + data-dictionary→harness (Q5/Q6) are affirmed. Six amendments fold in:

1. Q2 (hash vs slash) — REVISE toward HASH, ~5-2 (the council’s strongest pushback; ⚠ contradicts directing-authority rule 2). After full cross-talk a four-voice convergence — Gandon (Cool URIs/httpRange-14) + Baker (W3C Best Practice Recipes 3-vs-4) + Pandit (DPV principal author) + Davis (DA) — recommends Option H: a hash within the standard segment, https://w3id.org/opda/pdtf#Property, on a ground stronger than preference: rule 2’s slash override fails its OWN governing test. ODR-0004 §Consequences / ADR-0006 already fixed when slash earns its keep — “≥1,000 terms OR a named per-term-content-negotiation consumer” — and both limbs are unmet (~147 terms, no named consumer). The greenfield-cost argument shows a switch is cheap, not warranted. And the cited precedents are counter-evidence: DPV is slash-to-module + #-within (/dpv/pd#Address), hm is slash-to-module — both license hash for the core, neither supports flat-slash-per-term (which takes Recipe-4 dereference cost at Recipe-3 scale — the worst trade). The Q5 /harness/ split keeps the term vocabulary small (data-dictionary + profiles + instance data leave /pdtf/), which strengthens the hash case. Council recommendation: revert to the ODR-0004 hash now (Option H); adopt slash only when a reopening trigger fires, and then as slash-to-module + hash-within, never flat slash. Minority (Allemang on the schema.org flat-slash precedent; Knublauch deferring to the web-architects) holds for slash if opda commits to building per-term 303/content-negotiation (which w3id’s flat 302 does not provide). ⚠ DIRECTING-AUTHORITY CALL: rule 2 (no hash) was an explicit directive; the council’s reasoned majority is against it. The operator ratifies — accept Option H (hash), or override the council and keep slash with a committed per-term resolution build.
2. Q3 (versioning). Keep the version-free term IRI and carry both owl:versionIRI and owl:versionInfo on the ontology header (OWL 2 §3.1 — versionInfo alone has no machine identity), plus a dated release snapshot under /harness/release/<v>/.
3. Q4 (flat). Flat now, with a recorded reopening criterion: introduce a module segment only when a sub-vocabulary acquires an independent versioning clock OR is adopted/imported independently of the whole.
4. Q5 (split) — 7-0-0 (Davis fully withdrew). Keep /pdtf/ vs /harness/, governed by Baker’s binding placement procedure (the session’s most durable artefact, forced by Davis’s “standing tribunal” attack): (a) placement test — “would a consumer cite/version/deprecate this AS part of the normative standard?” Yes → /pdtf/, No → /harness/ (deterministic on all live cases + future kinds); (b) ambiguity fallback — default to /harness/ (the re-classifiable side), promote to /pdtf/ only by council, never demote a published /pdtf/ URI (DCMI: a published term URI is never reassigned); (c) one-directional dependency — nothing in /pdtf/ may rdfs:isDefinedBy/depend on /harness/. Realisation (settled): /harness/ is a PATH segment under ONE /opda/ PICG redirect, NOT a second coordinate namespace — the harness is definitionally opda-about-pdtf (co-stewarded, not separately delegable; Baker), and only a path segment is in the cold URI string at routing time (Allemang, httpRange-14). This is a REVISE of the proposal’s “two coordinate redirects” → one redirect, two paths. (Davis’s mechanical “published-graph + rdf:type” test aligns with ODR-0027 and is the membership heuristic behind the placement test.)
5. Q7 (SHACL shapes) — 7-0-0. Shape nodes (base + profile, opda: namespace) → /pdtf/ (co-located with their sh:targetClass; co-normative with the ontology). Shape documents (serialised files / named graphs) → /harness/. No /shacl/ namespace — the shapes-graph/data-graph separation (SHACL Rec §1.5) is a graph-level partition, not a naming one. (This resolves the directing-authority’s /shacl/ question: rejected.)
6. Q8 (profiles) — disaggregate + WARRANTY test. Profile documents + conformant-submission fixtures → /harness/. Profile shape-node/resource placement is governed by warranty, not subject-matter (Pandit): a profile opda governs and warrants as a normative PDTF conformance target → /pdtf/profiles/<form>; an overlay that merely maps to an externally-owned form (dct:source → third-party instrument: BASPI/TA6/CON29/LLC1) → /harness/profiles/<form>. Default the 31 form overlays to /harness/profiles/; promote individually on a recorded warranted adoption. ⚠ OPERATOR POLICY CALL (open): does opda publish the overlay profiles AS normative parts of the PDTF standard (→ promote to /pdtf/profiles/), or as opda’s mapping-views onto externally-owned forms (→ /harness/profiles/)? The council defaults to harness pending this ruling.

Framing correction (record): opda diverges from both DPV (slash-module + #-terms) and hm (/ns/pf/, /ns/sds/ context modules) on hash (Q2) and modules (Q4) — deliberately, on pre-publication + single-context grounds. DPV is a counter-precedent to flat-slash-per-term, not an endorsement of it; do not cite it as supporting slash.

Context and Problem Statement

ODR-0004 §Rules.1 (ratified at Council Session 004, 2026-05-27) commits the OPDA ontology to a single opda: HASH namespace but deliberately defers the literal base URI to the OPDA Working Group. The Council named three candidates (S004 Q7):

• https://opda.uk/ns/ — institutional, OPDA-org-lifecycle-dependent.
• https://w3id.org/opda/ — W3C Permanent Identifier Community Group (PICG); survives org changes.
• https://trust.propdata.org.uk/ontology/ — rejected at S004 on programme-namespace-coupling grounds.

Knublauch’s DA primary demand at S004 was specifically w3id.org/opda/ for the PICG persistence guarantee: w3id.org redirects are maintained by the W3C community independent of any single organisation’s lifecycle. Pandit cited the same reason as principal author of DPV’s adoption of https://w3id.org/dpv/.

The WG decision was taken on 2026-05-27: adopt w3id.org/opda/ with hash suffix, i.e. terms resolve as https://w3id.org/opda/#Property, https://w3id.org/opda/#LegalEstate, etc.

w3id.org does not itself host content — it issues HTTP redirects to the actual ontology hosting. The infrastructure work this ADR records is therefore distinct from the modelling decision: it commits OPDA to (a) submitting the redirect rule to the W3C PICG repository, (b) hosting the actual ontology artefacts at a stable target the redirect points at, and (c) maintaining the redirect-target pairing across future moves.

This is the first cross-corpus ADR connecting the ADR programme (engineering decisions) to the ODR programme (ontology-modelling decisions). DCAP boundary: ODR-0004 owns the namespace string; this ADR owns the redirect infrastructure.

Decision Drivers

• PICG persistence guarantee. URIs stay stable independent of OPDA’s organisational survival, rebrand, or domain changes — the costly-to-fix problem (DPV slash→hash precedent: six ecosystem-months to propagate).
• DPV precedent. Pandit’s choice for DPV (w3id.org/dpv/) is the closest analogue and the principal-author endorsement.
• OPDA already controls openpropdata.org.uk. The institutional web domain exists and can serve as the redirect target without new DNS work.
• w3id.org submission is one PR. Onboarding cost is small; ongoing cost is near-zero (only edit the redirect target when hosting moves).
• Single point of failure mitigation. PICG-maintained redirects are infrastructure OPDA does not have to operate.

Considered Options

• A — https://opda.uk/ns/ (institutional only). Direct, no external dependency. Rejected because OPDA-org persistence is weaker than W3C PICG.
• B — https://w3id.org/opda/# via W3C PICG redirect (chosen). PICG persistence; redirect to OPDA-hosted TTL.
• C — https://openpropdata.org.uk/ns/# (institutional, opda’s actual domain). Briefly chosen 2026-05-27 then reverted in favour of B on the persistence-guarantee argument.
• D — https://trust.propdata.org.uk/ontology/. Rejected at S004 on programme-namespace-coupling grounds (no champion).

Decision Outcome

Chosen option: B — https://w3id.org/opda/# via W3C PICG redirect, because it satisfies all four decision drivers and meets Knublauch’s S004 DA primary demand. The redirect target is https://openpropdata.org.uk/ontology/ — the OPDA institutional domain serves the actual TTL; w3id.org issues the persistent URIs that consumers cite.

The literal prefix declaration in all OPDA TTL artefacts will be:

@prefix opda: <https://w3id.org/opda/#> .

The redirect rule submitted to perma-id/w3id.org is shaped:

# /opda/.htaccess (or equivalent)
RewriteRule ^(.*)$ https://openpropdata.org.uk/ontology/$1 [R=302,L]

(302 not 301 — leaves the redirect editable without breaking cached consumers. Switch to 301 once the hosting target stabilises.)

The hosting target itself (https://openpropdata.org.uk/ontology/) is provisioned by the Astro site build (see ADR-0003); the generator emits opda-classes.ttl, opda-shapes.ttl, and opda-annotations.ttl into a public/ontology/ tree per ODR-0004 §6a.

Consequences

• Good, because URIs stay persistent independent of OPDA’s organisational lifecycle (W3C PICG persistence guarantee — DPV precedent).
• Good, because Knublauch DA primary demand at S004 Q7 is met; ODR-0004 can move status: proposed → accepted once the literal string lands in §Rules.1.
• Good, because OPDA-side maintenance is reduced to a single redirect line in a public W3C-managed repository.
• Good, because the redirect target (openpropdata.org.uk/ontology/) is on a domain OPDA already controls — no new DNS work, no new domain registration.
• Good, because the choice of redirect target is decoupled from URI policy: future moves (e.g. content-delivery-network hosting) are deploy-side concerns and do not invalidate any minted URI.
• Bad, because resolution now depends on W3C PICG infrastructure; a w3id.org outage breaks all OPDA URI dereferencing for the duration.
• Bad, because the PR to perma-id/w3id.org is asynchronous: merge time is community-paced (typically days to a week). Mitigation: TTL emission can proceed in parallel; consumers loading local copies are unaffected.
• Bad, because the redirect must be maintained in two places (the perma-id PR + the OPDA-side hosting target). Mitigation: redirect target rarely changes; OPDA-side hosting is part of the existing Astro build pipeline.
• Neutral, because the hash-vs-slash decision (ODR-0004’s hash commitment) is preserved — terms are addressed as fragments of the ontology document, not per-term URIs. Reopening trigger stays as ODR-0004 §Consequences records it.

Confirmation

The ADR is honoured when all five hold:

1. PR submitted. A pull request against perma-id/w3id.org adding /opda/ redirect to https://openpropdata.org.uk/ontology/.
2. PR merged. W3C PICG community-merge completes (typically 2-10 days).
3. Live resolution. curl -IL https://w3id.org/opda/foundation.ttl returns 302 → 200 with Content-Type: text/turtle.
4. ODR-0004 amended. §Rules.1 placeholder (“WG-owned open question”) replaced with the literal string https://w3id.org/opda/#; §References cites this ADR.
5. Status sweep complete. Every ODR with depends-on: [ODR-0004] moves status: proposed → accepted per the dependency chain (ODR-0002 through ODR-0018 inclusive).

CI test (deploy pipeline): byte-identity of regenerated TTL against committed sources (ODR-0004 §6a) plus a curl probe asserting both https://w3id.org/opda/foundation.ttl and https://openpropdata.org.uk/ontology/foundation.ttl return the same byte-identical document.

Pros and Cons of the Options

A — https://opda.uk/ns/ (institutional only)

• Good, because direct: one URL, no external dependency, full OPDA control.
• Bad, because OPDA must commit to keeping opda.uk resolving forever, including across organisational changes — Knublauch’s S004 primary attack.
• Bad, because no DPV-style precedent; pure institutional persistence is weaker than W3C PICG.

B — https://w3id.org/opda/# via W3C PICG redirect (chosen)

• Good, because PICG persistence guarantee survives OPDA org changes (the costly-to-fix problem).
• Good, because aligns with DPV’s choice (Pandit’s principal-author endorsement).
• Good, because OPDA-side maintenance cost lowered to a single redirect line.
• Bad, because resolution depends on w3id.org availability (a W3C-community-operated service).
• Bad, because PR-to-merge cycle is community-paced — short asynchronous delay before resolution goes live.

C — https://openpropdata.org.uk/ns/# (OPDA institutional, full domain)

• Good, because uses OPDA’s actual existing domain (openpropdata.org.uk) rather than the opda.uk shortened form.
• Bad, because still institutional persistence — same critique as Option A.
• Bad, because no W3C PICG layer of indirection; future moves invalidate URIs.

D — https://trust.propdata.org.uk/ontology/ (programme-namespace)

• Bad, because couples the namespace string to a programme name (trust.propdata) — rejected at S004 on lack-of-champion grounds.

More Information

• Modelling decision (ODR corpus, cross-corpus dependency): ODR-0004 — PDTF Ontology Foundation, §Rules.1 (single opda: HASH namespace) and §References (Open questions: namespace string). This ADR records the engineering realisation of that rule.
• Ratification provenance (ODR corpus): Session 004, Q7 (namespace-as-blocker; w3id.org alternative named by Knublauch). Vote 9-0 with full Knublauch DA withdrawal on all four primary attacks.
• Author-only Session 003b (planned): records the WG namespace decision into ODR-0004’s ## Rules and the adoption record track record. Pairs with this ADR.
• DPV precedent: W3C Data Privacy Vocabulary at https://w3id.org/dpv/ — Pandit’s principal-author choice, cited by Knublauch as the operationally-strongest analogue. DPV’s slash→hash transition in 2019 took approximately six ecosystem-months to propagate; the persistence guarantee is what makes that recoverable.
• W3C PICG repository: https://github.com/perma-id/w3id.org. Submission process: open a PR adding the /opda/ directory with redirect rules. Community review by PICG maintainers.
• Hosting target: https://openpropdata.org.uk/ontology/ — see ADR-0003 for the Astro build pipeline that emits and serves the TTL artefacts.
• Reopening trigger (inherited from ODR-0004 §Consequences): the hash decision is reversible only at high cost. The WG SHOULD record a concrete reopening criterion (suggested: any single ontology file exceeds 1,000 terms in active dereference traffic OR a named consumer requests per-term content negotiation).
• First cross-corpus ADR. This ADR is the bootstrap of the ontology-implementation ADR programme referenced in docs/plan/council-followup-sessions.md §1 (“Implementation is handled by a separate ADR programme”). Subsequent ADRs will realise additional ratified ODRs.

Amendments

• 2026-09-07 — Namespace scope separated. ODR-0065 records the directing authority’s new-model namespace selection. This ADR’s definitive block remains authoritative for the schema-derived corpus only; its identifiers are unchanged.

• 2026-08-22 — Chair-authority terminology correction without RDF reminting. Maria Harris, OPDA Chair, clarified that the inherited technical material was not an OPDA-endorsed predecessor standard or scheme. The unchanged /pdtf/** IRIs are the stable identifiers of the ontology derived from the PDTF schema. Their path text is historical identity, not semantic authority and not an SPDTF generation or version. The earlier “PDTF standard” and “normative standard” wording retained in the dated 2026-06-01/02 record is historical decision provenance; this amendment and the current identifier-scheme block govern its present interpretation. No class, property, concept, shape, graph or release IRI is changed.