Skip to content

ADR-0021: Core Placement Is Source-Classified, Existentially Quantified — Not Cross-Source Popularity

  • Status: Accepted
  • Date: 2026-07-05

Context: Why This Review Happened

ADR-0020 defined "core" for XwkOnt (the base module a project's own vocabulary is built from, as distinct from optional extension modules) but explicitly declined to decide which of the 17 currently-reviewed concepts actually belong there — that placement criterion was left as separate, not-yet-started work. This ADR is that criterion.

The first framing considered was, again, judgment-based: XwkOnt evaluates each concept's structural importance or cross-source recurrence and sorts it into core or extension by that evaluation. The maintainer rejected this approach a second time, more sharply: XwkOnt should not apply its own judgment about whether a concept is "core" at all. If a source ontology classifies a concept as part of its own core/base module, that classification stands on its own — it is not negated by the fact that other sources don't share it, or that it isn't "widely adopted" across the crosswalk. Wide adoption is a social fact, not a validity criterion (the maintainer's own example: "the world is flat" was widely accepted, and being widely accepted didn't make it true — the inverse holds too: a concept isn't disqualified from being core just because most sources don't happen to treat it that way).

Deliberation

1. This isn't a new principle invented for this decision — it's already documented, in three places. docs/ARCHITECTURAL_PRINCIPLES.md already states: - Principle 4 — Source First: "Authoritative sources remain authoritative. XwkOnt links to, cites, and compares them." - Principle 6 — Neutrality: "XwkOnt documents and compares. It does not advocate or replace existing foundational ontologies." - Principle 7 — Open-World Philosophy: "The project assumes knowledge evolves."

Requiring cross-source agreement before honoring one source's own core-module classification would itself be an act of advocacy — XwkOnt asserting "this source's internal structure doesn't count until enough others agree with it," which is exactly the kind of correctness judgment Neutrality rules out, and exactly the "widely-accepted = valid" conflation Open-World Philosophy is written to avoid.

2. The quantifier has to be existential (∃), not majority-based, once Source First is taken seriously. If a concept is real evidence against simply requiring "most sources agree it's core," then the honest reading of Source First is: one source's own core-module classification is sufficient on its own. Requiring two, or a majority, would be an arbitrary threshold with no more principled grounding than requiring zero (i.e., XwkOnt's own judgment) — it just moves the same problem to a different number.

3. Tested directly against the concept this whole investigation started from. Situation/State of Affairs (situation-state-of-affairs.md) is the sharpest available test case: GFO's Configuration/Situation classes live in gfo-base.owl — GFO's own core module (per ADR-0020's evidence). DOLCE's real treatment lives in its DnS extension. UFO's lives only in the OntoUML vocabulary, this project's own already-established secondary/derived-artifact tier for UFO (used this way across role.md, universal-type.md, and situation-state-of-affairs.md itself). Under this ADR's rule, Situation is core — full stop — because GFO alone places it in GFO's own core, regardless of what DOLCE and UFO do with it internally.

4. The consequence flagged when this was first discussed (nearly all 17 reviewed concepts will likely qualify as core) is not a defect — it is the correct result of taking Source First seriously, not a sign the rule is too loose. Most crosswalk correspondences built so far were sourced from a source's own core/base file to begin with, precisely because that is where each source's foundational vocabulary typically lives. A rule that produced a small, hand-picked "core" subset would be the one requiring an undocumented judgment call — this rule doesn't, and that it turns out permissive is a consequence of the evidence, not a design flaw.

5. This does not touch how trustworthy any individual correspondence is judged to be. Per the maintainer's own framing, "concept usage can be determined by the user in the concept's crosswalk" — a concept's placement in XwkOnt-core only answers "does this belong in the base module," not "how much should a given ontologist trust this specific correspondence." That judgment already lives in each crosswalk's own confidence values and Semantic Comparison Notes, untouched by this ADR.

What Counts as a Source's Own "Core" — Verified So Far

Source Core artifact (verified) Extension artifact(s) (verified) Status
BFO bfo-core.ttl (21838-2/owl/, BFO-ontology/BFO-2020) Separate relation-ontology / IAO extension files Modularized, directly verified (ADR-0020)
DOLCE DOLCE-Lite (2009-relabeled "DOLCE-CORE") DnS (ExtendedDnS.owl) Modularized, directly verified (ADR-0020, this session's situation-state-of-affairs.md pass)
GFO modules/gfo-base.owl modules/situation/*.ttl; the parallel gfo-core: namespace found inside the situation module (GFO's own README: "parallel extensions... not all... consistent") Modularized, directly verified (ADR-0020, this session)
UFO The 2021 comprehensive paper (Guizzardi et al.) — UFO's own primary theoretical exposition The OntoUML vocabulary artifact — a derived modeling-notation artifact, already treated as secondary/fallback evidence throughout this repository's existing crosswalks Not file-modularized the way BFO/DOLCE/GFO are, but already artifact-tiered by this project's own established practice — reused here, not newly invented
SUMO, YAMATO, TUpper, GUM Not checked in this pass Not checked in this pass Unresolved — only matters for a concept whose sole supporting source is one of these four; not yet a blocker for any of the 17 concepts, all of which have at least one BFO/DOLCE/GFO/UFO correspondence

Decision

A concept belongs in XwkOnt's core module (xwkont-core:, data/ontology/core.ttl) if at least one contributing source ontology's correspondence for that concept is drawn from that source's own core/base-module artifact — existentially quantified, not majority-quantified. No cross-source popularity, adoption-count, or structural-importance filter is applied on top of this. A source placing a concept in its own extension tier does not count against that concept's core status if any other contributing source places it in that source's own core tier; a concept whose every contributing source locates it only in that source's own extension tier does not qualify for XwkOnt-core under this rule (it may still be a real, reviewed Concept — placement and review status are independent, per ADR-0020's own scope).

This is a placement criterion, not a re-review — it does not change any crosswalk's editorial status, confidence values, or mapping assertions. It only answers the single question ADR-0020 deferred: does this concept go in the base module or an extension module.

Scope

This ADR does not:

  • Apply this criterion to the 17 currently-reviewed concepts and sort them — that remains a separate, mechanical (not judgment-based) follow-on pass: for each concept, check whether any of its Source Ontology Correspondences rows cites a source's own core/base artifact (largely already recorded in each crosswalk's existing Reference column) against the table above.
  • Resolve SUMO/YAMATO/TUpper/GUM's own internal core/extension modularization (if any) — deferred until a concept's placement actually depends on one of those four sources alone.
  • Decide the physical file/directory/namespace structure of XwkOnt's own future extension module(s) — still separate implementation work per ADR-0020.
  • Revisit the original 8 0.1.0 concepts already in core.ttl — no re-verification is required or proposed; nothing in this ADR casts doubt on their existing placement.

Consequences

Positive

  • Closes ADR-0020's deferred placement question with a rule that requires zero XwkOnt-side structural judgment, fully consistent with Source First, Neutrality, and Open-World Philosophy as already documented — not a new principle invented for this decision.
  • Directly testable against evidence already collected: most of the per-source core/extension classification work for BFO, DOLCE, GFO, and UFO was already done incidentally across this session's situation-state-of-affairs.md investigation and ADR-0020 itself.
  • Avoids an unprincipled numeric threshold (why 2 sources and not 3, or a majority) that a stricter quantifier would have required defending.

Trade-offs

  • Likely results in most or all of the 17 reviewed concepts qualifying as XwkOnt-core, leaving the extension tier sparsely populated (or empty) until weaker-evidence future batches (per the internal version-roadmap's queued, 1–2-source-threshold candidates) are drafted and reviewed — the tier this ADR makes room for may not see real use for some time.
  • SUMO/YAMATO/TUpper/GUM's own core/extension structure (if any) remains unresolved; a future concept whose only source is one of these four will need that gap closed before this rule can be applied to it.

References

  • docs/adr/ADR-0020-define-core-as-base-module-not-domain-tier.md (extended, not reopened)
  • docs/ARCHITECTURAL_PRINCIPLES.md (Principle 4 — Source First; Principle 6 — Neutrality; Principle 7 — Open-World Philosophy)
  • docs/crosswalks/concepts/situation-state-of-affairs.md (the test case: GFO core-classified, DOLCE/UFO extension-classified)
  • docs/references/ref-gfo.md, docs/references/ref-dolce-borgo-2022.md, docs/references/ref-bfo-2020.md, docs/references/ref-dolce-extendeddns-owl.md (per-source core/extension evidence)