Single canonical UFO typing axis

Status note. Draft, pending ratification (directing authority = acting ratifier). Council session-049 deliberated this; its verdict — adopt the single-axis invariant, NOT a corpus-wide re-typing — matches hm ODR-0100 and stands.

Context and Problem Statement

OPDA types classes ontologically in a partial, mixed way: gUFO typing gated to a few descriptive Quality leaves (ADR-0034/0045); Roles/Relators via opda:Role/RoleMixin/Relator meta-classes; OntoClean meta-property markup (ADR-0046). The proposition asked: adopt a single canonical UFO typing axis corpus-wide to make OntoClean rigidity decidable and “unlock an anti-rigid-subsumption CI gate” — vs keep the gated/partial typing.

Council session-049 found the proposition’s headline was stale and its expensive demand wrong, and converged (4–0) on a smaller, correct disposition. This matches prior art hm ODR-0100 (single canonical UFO typing axis) + ODR-0092 (layer/stereotype SKOS schemes).

Decision Drivers

• Decidable OntoClean rigidity — the precondition for any anti-rigid-subsumption gate (checker totality, not reasoner decidability).
• Single source of truth — one axis for a class’s ontological nature; any second axis a derived view.
• Cost / YAGNI — corpus-wide re-typing manufactures unconsumed axioms (ODR-0101 “a latent liability, not a free asset”).
• Honesty — keep the upper-ontology anchor inert (no entailment, no unmet conformance claim).

Considered Options

• (A) Corpus-wide rigidity-vector re-typing of all ~41 classes — REJECTED 4–0 (manufactures ~31 unconsumed axioms; coerces ±R onto non-sortal categories).
• (B) Keep gated/partial typing, do nothing — insufficient (leaves the single-valued-ness invariant unstated/unenforced + a real emitter defect, below).
• (C) Adopt the single-axis invariant + gates (hm ODR-0100) — CHOSEN. State the invariant, enforce it, fix the derivation; do NOT re-type.

Decision Outcome

REVISE → adopt the single-axis invariant, not corpus-wide re-typing (Council session-049, 4–0; matches hm ODR-0100). Two stale premises were rejected: (i) the anti-rigid-subsumption gate “to unlock” already ships green (ADR-0046 checks 8/9, TBox-only); (ii) the single category axis (opda:ufoCategory, owl:AnnotationProperty, closed UFOCategoryScheme) already exists on 39/40 classes (opda:SpecialCategoryScheme, a skos:ConceptScheme, correctly exempt — scope the invariant to first-order universals, not “endurants”).

Adopt (all atomic — ships green or not at all):

1. State the invariant — ufoCategory is the single canonical typing axis (scoped to first-order universals; scheme/meta artefacts exempt); ontoCleanRigidity/Identity/Dependence are a derived view (a function of ufoCategory), never an independent assertion; the gufo: rdf:type alignment (ADR-0034) is a derived view via skos:closeMatch, not a second axis.
2. Axis-consistency meta-shape over the 8 rigidity-tagged classes — fails CI if a class’s rigidity/identity contradicts its ufoCategory signature.
3. Edge-targeted growth-frontier guard — fails CI when an intra-opda: rdfs:subClassOf edge has an endpoint lacking a rigidity projection (catches the silent-false-green at edge-authoring time, without pre-typing edgeless classes).

Recommended (not a soundness precondition): automate g(ufoCategory) to replace the hand-table (fixes the lying docstring; makes single-valued-ness hold by construction). Reuse ufoCategory/UFOCategoryScheme; do NOT mint a parallel scheme.

Decline: corpus-wide rigidity-vector population (~31 unread tags); any minted parallel scheme; ±R coercion on non-sortal categories (Information Object, Quality, Quality Value, Collective).

Consequences

• Good — makes single-valued ontological typing an enforced, prospective invariant (sound for class N+1), at near-zero cost; matches hm ODR-0100.
• Good — surfaces and fixes a real generator defect (below).
• Neutral — the ~31 untyped-for-rigidity classes stay untyped (by design).
• Bad — adds 2–3 CI gates + an emitter change to maintain.

Confirmation

• The axis-consistency + edge-targeted meta-shapes ship green, atomically, over the tagged set; full make ci; byte-identity.

Council sessions and findings

• session-049 (Council B — Queen Guizzardi, DA Davis, panel Guarino + Allemang; working/session-049/): REVISE 4–0 — adopt the invariant + gates, reject corpus-wide re-typing. The formal core (Guarino): rigidity must denote a function Type → {+R,−R,~R,non-R}; two unconstrained parallel axes make it a relation → the SHACL gate’s −R⋣+R check becomes ill-typed (checker totality, inside OPDA’s gate — not reasoner decidability).
• VERIFIED BUG (Allemang’s find, Guarino+Davis confirmed, Queen re-verified): _ONTOCLEAN_TAGS (tools/opda-gen/src/opda_gen/emitters/ufo_categories.py:298-315) is a hand-maintained 8-row dict keyed by class local-name; the _RIGIDITY_COMMENT/_IDENTITY_COMMENT (:323/:334) claim “Derived from each type’s opda:ufoCategory signature” but nothing computes it. The 8 values are correct today (gate green) — so the docstring is true in extension, false in mechanism: a latent session-047-class defect that bites when a sortal subkind edge is added through an untyped middle class. Fix: automate g(ufoCategory) (the 9 UFOCategoryScheme signatures already hold the table as data) — REQUIRED is the edge-targeted guard; RECOMMENDED is the automation.
• Empirical anchors (verified at file:line): ADR-0046 checks 8/9 green (three_graph_test.py:348/:417); ufoCategory on 39/40 classes; ontoCleanRigidity on exactly 8 domain classes; only 5 intra-opda: rdfs:subClassOf edges (lattice flat by ODR-0027; all other supers external).
• Held dissent — Davis (DA), non-blocking: the rigidity projection stays edge-participant-scoped (never on a class in no intra-opda: subsumption edge); ships green-or-nothing. Re-open trigger: a rigidity vector authored on an edgeless class, OR the rule shipped as ODR prose without a consuming gate → flat REJECT.

Outstanding work and open tasks

ALL DONE (2026-06-17, commits 4589bb3 + the records): implemented, verified, full make ci green. Items 1–7 below all landed — the invariant is stated (this ADR §Decision Outcome); g(ufoCategory) replaces the hand-table (values derived, docstring no longer lies; byte-identity held — the emitted 8 tags unchanged, proven by the .py-only commit); the axis-consistency check (three-graph check 10) + edge-targeted growth-frontier guard (check 11) ship green with positive controls; SpecialCategoryScheme exemption recorded; session-049 record + adoption.md row written. The corpus-wide rigidity-vector population stays REJECTED (Davis’s edge-participant scope held).

1. State the single-axis invariant (scoped to first-order universals) in the doctrine ODR / ADR.
2. Add the axis-consistency meta-shape over the 8 tagged classes.
3. Add the edge-targeted growth-frontier guard (required for soundness).
4. Automate g(ufoCategory) in the emitter — replace the _ONTOCLEAN_TAGS hand-table (recommended; fixes the lying docstring).
5. Confirm SpecialCategoryScheme exemption is recorded (first-order-universals scoping).
6. Proof obligations: gates ship green atomically; byte-identity; full make ci.
7. Write the session-049 record + adoption.md track-record row.

More Information

• Working brief: docs/HANDOVER-2026-06-17-entailment-doctrine-domain-range-as-signal-relationship-layer-rework.md (Part C5; Part E decision 3).
• Related: ADR-0034 (gUFO typing), ADR-0045 (ufocategory/gUFO scheme), ADR-0046 (OntoClean meta-property markup); ODR-0027 (classification-over-inheritance).
• Prior art adopted (cite here only; strip from normative ODR text): ~/source/hm/semantic-modelling ODR-0100 (single canonical UFO typing axis), ODR-0092 (layer/stereotype schemes), ODR-0101 (BFO backbone kept inert).

Amendments

• 2026-06-17 (a) — created in Draft; decision OPEN pending Council B.
• 2026-06-17 (b) — Council B (session-049): REVISE 4–0 → adopt the single-axis invariant + 2 gates + the emitter-derivation fix; REJECT corpus-wide re-typing; matches hm ODR-0100. Davis held dissent (edge-scope, non-blocking) recorded.