Relationship-residue completion — events, information objects, verification, and the aboutness/provenance boundary

Context and Problem Statement

ODR-0032 (council session-047) built the relationship spine (the agent/transaction/role/property object properties) and the three-disposition residue register (GATED / VALUE-SLOT / RESIDUE-PENDING). It did not adjudicate the perdurant events (Milestone, LeaseExtensionEvent, UPRNSuccessionEvent, NameChangeEvent), the Information Objects (Survey, Search, Comparable, NearbyFacility, AttachedDocument), VerificationActivity, LeaseTerm, or the chain predicates — session-047 had no PROV-O voice, and these classes are overwhelmingly prov:Activity/prov:Entity.

A 2026-06-30 review of the /ontology/classes diagrams surfaced the consequence: 18 of 40 classes render disconnected, because the diagram draws only object-property class→class edges. The review is the trigger; the residue is the substance.

The decisive corpus fact (convener-verified). opda:partOfTransaction (Milestone→Transaction), opda:concerns (Transaction→LegalEstate), opda:dependsOnTransaction, opda:chainMembers, opda:chainStatus, opda:appliesTo, opda:updatesRegistryRecord are used in committed diagnostic exemplars but none is declared owl:ObjectProperty in the gated TBox — the ADR-0048 §As-built “undeclared orphan predicates, left as-is.” So the question is not “mint new relations” but “declare and gate edges the corpus already commits in its ABox,” and the chain exemplar exists (the session-047 “no exemplar yet” deferral is stale). The five info-object subject classes appear in zero exemplars.

Decision Drivers

• Faithfulness without over-reification. The §R1 endpoint-IC test (both endpoints +I AND a worked competency query) governs; source containment of a −I quality is a value-path, not an edge.
• Reuse over mint (parsimony). Where PROV-O already names the relation, declaring a parallel opda: predicate duplicates PROV and orphans the existing triples.
• Aboutness ≠ provenance. PROV-O expresses derivation, influence, and attribution — not topical subject-hood. Substituting prov:wasDerivedFrom for “is about” is a false provenance claim.
• No undeclared edges. An object property used in the ABox but undeclared in the TBox is ungoverned (invisible to the coverage gate, no SHACL value-shape) — the founds/mediates rangeless defect in another guise.
• No collapse-by-silence. A warranted-but-unexercised relation is registered (RESIDUE-PENDING), never dropped.

Considered Options

• Option A (chosen) — Adjudicate each residue edge by §R1 into one disposition; declare+gate the exemplar-attested edges, reuse PROV where it carries the relation, register the warranted-but-unexercised ones, value-slot the −I endpoints.
• Option B — Gate every candidate to “connect” the diagram. Rejected: over-reifies −I endpoints (LeaseTerm), mints PROV duplicates (VerificationActivity), and gates predicates no exemplar exercises (info-objects) — the §R1(b) violation. The diagram disconnection is fixed by rendering rdfs:subClassOf (ADR-0055), not by minting.
• Option C — Mint bespoke opda: predicates for the PROV-native relations (opda:usesEvidence, opda:wasInformedBy). Rejected: duplicates PROV-O, contradicts the reuse driver, and orphans the exemplars’ existing prov: triples.
• Option D — Leave the undeclared exemplar predicates as-is. Rejected: they are ungoverned (ADR-0048 flagged them for “a future council” — this is it).

Decision Outcome

Chosen option A. The residue resolves under one rule — declare-and-gate where the corpus already commits the edge; reuse PROV where it names the relation; register-and-defer where the relation is warranted but unexercised; value-slot the −I endpoints — with full Council convergence and no held-as-live dissent (session-051; DA Davis withdrew on the gated edges, held with register-satisfied triggers elsewhere).

Consequences

• Good, because the two exemplar-attested orphan predicates become governed, type-pinned, navigable edges (Milestone→Transaction, Transaction→LegalEstate) instead of ungoverned ABox predicates.
• Good, because the PROV-native relations (verification, comparable-informs-valuation, event-revision/derivation) are surfaced as the navigable edges they already are — the model is not disconnected, the diagram was.
• Good, because the aboutness/provenance boundary is now doctrine (§R5): a future “Survey about Property” edge will be opda:aboutProperty, not a mis-stated prov:wasDerivedFrom.
• Bad, because it is a byte-identity-affecting regeneration (two new object-property declarations + SHACL shapes + two committed competency queries + regenerated ontology-model.json/graph/expected-reports) — delegated to ADR-0056.
• Neutral, because the participant-attribute backlog and the W3C-Org-vs-bespoke Kind-layer choice remain out of scope (ODR-0006/0032).

Confirmation

Compliance is verified by the ci-object-property-coverage gate (ADR-0048, extended by ADR-0056): the two GATED edges are declared, type-pinned (documentary rdfs:domain/rdfs:range per ODR-0032 §R1 as amended by session-050, never entailed — ADR-0035 zero-domain/range-triple proof — plus a SHACL sh:class value-shape + subject-guard), and each carries one worked SPARQL competency query; every RESIDUE-PENDING register entry carries a named disposition + reason + auto-gate condition (an empty/“TODO” entry fails the gate); no PROV-native relation is shadowed by a bespoke opda: predicate. The diagnostic exemplars validate against the new SHACL shapes.

Rules

R0 — Endpoint meta-categories and identity criteria (per-kind discipline, ODR-0001 §“What an ODR records”)

The classes this ODR newly edge-connects, with their UFO/DOLCE meta-category and IC over the named hard cases:

• opda:Milestone — UFO Event (DOLCE perdurant; Guizzardi et al., ER 2013 §3). IC: identified by its participation structure + temporal extent within its Transaction; two milestones of the same phase-type in one transaction are distinct iff their temporal extents differ. Hard case: a re-issued completion milestone after a fall-through is a new Event (new temporal extent), not the same milestone re-dated.
• opda:Transaction — UFO Relator (ODR-0005/0007), well-founded on the Property/Proprietor/participant Kinds it mediates. IC: bounded-context identity (ODR-0005, 12–0) — a Transaction is not owl:sameAs-collapsible across registries.
• opda:LegalEstate / opda:RegisteredTitle — UFO Substance-Kind / non-physical endurant (ODR-0005; DOLCE D18 §4.2). IC: the estate’s identity persists through registry record changes; a re-registration under a new title number is the same LegalEstate (the recordsEstate arm), a distinct RegisteredTitle.
• opda:VerificationActivity — UFO Event (prov:Activity). IC: participation structure (who verified, what evidence used, when). Hard case: a re-verification is a new VerificationActivity (opda:hasVerificationSuccessionStatus), not a mutation of the prior one.
• Information Objects (opda:Survey/Search/Comparable/NearbyFacility/AttachedDocument) — UFO Information Object, non-physical endurant with extrinsic IC ⟨issuing authority, reference, issue date⟩ (ODR-0008d). Reaffirmed here as legitimate +I edge endpoints — but see §R3/§R4.

R1 — GATED: declare the exemplar-attested orphan predicates

The following edges are GATED — declared owl:ObjectProperty, documentary rdfs:domain/rdfs:range (ODR-0032 §R1 / S050; AI-signal, never entailed), a SHACL sh:class value-shape + relator/event subject-guard (ODR-0013), and one committed worked SPARQL competency query each:

Both predicates already occur on instances in exemplars/simple-transaction-with-milestones.ttl and exemplars/chain-of-transactions.ttl; this rule declares and governs them, it does not mint them. opda:concerns ≠ opda:concernsProperty: the former conveys the legal interest (LegalEstate), the latter the physical Property (ADR-0048 §As-built named both).

R2 — USE-EXISTING-PROV: the relation is PROV-native, no opda: mint

These relations are carried by existing PROV-O properties and MUST NOT be shadowed by a bespoke opda: predicate. They are emitted on the exemplars (where absent, per ODR-0009’s design) and surfaced by the diagram’s PROV layer (ADR-0055); the model is navigable through them:

• VerificationActivity → Evidence / AttachedDocument: prov:used (PROV-O §2). VerificationActivity → verifier Agent: prov:qualifiedAttribution / prov:wasAssociatedWith (PROV-O §3). Realises the ODR-0009-designed PROV backbone (§“PROV-O mapping”).
• Comparable → Valuation: prov:wasInformedBy (ODR-0008d Rule 3 — already designed).
• NameChangeEvent → Person: prov:wasRevisionOf / prov:wasAssociatedWith. UPRNSuccessionEvent → prior Address/Property: prov:wasDerivedFrom (ODR-0005 §6a).

opda:AttachedDocument’s evidentiary tie remains role-via-opda:evidenceType (ODR-0027 §R6 / ODR-0024 R7 — “not evidence by mere attachment”); no new edge.

R3 — RESIDUE-PENDING: warranted, but bar (b) unmet — register, do not gate

Added to the ODR-0032 relationship-residue register with a named auto-gate condition:

R4 — VALUE-SLOT: −I endpoints carry no inter-entity edge

• opda:LeaseTerm (rdfs:subClassOf time:ProperInterval) — a DOLCE temporal/quality-region (−I; D18 §4.2); the opda:leaseTerm join is a datatype/value path (time:hasBeginning/hasEnd), not an object property. (Unanimous, 7–0.)
• The name in opda:NameChangeEvent — −I; reaffirms ODR-0032 §R2’s struck hasName (string-literal datatype; the Person tie is prov:wasRevisionOf).
• NearbyFacility proximity — opda:distanceInMiles is a −I quality; only an explicit opda:aboutProperty “near Property P” edge is RESIDUE-PENDING (§R3), the distance itself is a value-slot.

R5 — Aboutness is not provenance (doctrine)

PROV-O carries derivation (prov:wasDerivedFrom), influence (prov:wasInformedBy), generation (prov:wasGeneratedBy), attribution (prov:qualifiedAttribution), and specialization (prov:specializationOf, entity→more-general-entity) — it does not carry topical subject-hood (“this Information Object is about subject X”); PROV-DM §5 has no topic/isAbout term. Therefore:

• A topical aboutness relation (e.g. a Survey about a Property) is modelled as a bespoke opda: predicate (opda:aboutProperty), not as prov:wasDerivedFrom/wasInformedBy — substituting a provenance term for aboutness asserts a false provenance claim (a Survey is generated by an inspection Activity, not derived from the Property).
• Conversely, a genuine derivation/influence/attribution relation MUST reuse the PROV term (§R2), not a parallel opda: predicate.

This is the boundary that resolves the session’s central split (Moreau, conceded by Kendall): reuse PROV for provenance; mint opda: for aboutness; never conflate the two.

More Information

• Council: session-051 (Full Council; Queen Kendall, DA Davis; Moreau/Guizzardi/Guarino/Allemang/Cagle). Full convergence, no held-as-live dissent.
• Realises / extends: ODR-0032 §R1 endpoint-IC test + residue register (this ODR adds §R3 entries and the §R5 aboutness/provenance boundary).
• Companion engineering record: ADR-0056 (declare partOfTransaction/concerns, the SHACL shapes + competency queries, the residue-register + ci-object-property-coverage extension, regeneration). The diagram realisation (rdfs:subClassOf render layer + the shipped cross-section links) is ADR-0055.
• Source designs: ODR-0009 (the VerificationActivity PROV backbone), ODR-0008d (Information Objects + prov:wasInformedBy), ODR-0007 (chain).
• Governance: per the OPDA handoff, this record stays proposed until the Modelling Sub-Committee draft-adopts; accepted only after AGM ratification.
• Out of scope: the participant-attribute backlog; the W3C-Org-vs-bespoke Kind-layer choice (ODR-0006 §Freeze-gate(b)).