Descriptive/Directive ICE definitions: what relation does 'consists of' express?

[DavidGLimbaugh]

Four CCO definitions build Information Content Entities out of pluralities:

• Descriptive Information Content Entity (ont00000853): "An Information Content Entity that consists of a set of propositions that describe some Entity."
• Directive Information Content Entity (ont00000965): "An Information Content Entity that consists of a set of propositions or images (as in the case of a blueprint) that prescribe some Entity."
• Religion (cco:ont00000616) and Ideology (cco:ont00000696): "...consists of a collection of claims..."

I would like to know what relation 'consists of' expresses here. Three candidates occur to me. None of them works, and the failures are instructive enough that I think a small documentation fix could spare future readers the exercise.

1. Set membership. Then the definitions quantify over sets, and BFO 2020 (ISO/IEC 21838-2) admits no sets or other abstracta. An ICE is a generically dependent continuant (obo:BFO_0000031), and BFO already tells us what that is: "the pattern or content which b shares at t with actual or potential copies" (definition of concretizes, obo:BFO_0000059). On the set reading, three core branches of the ICE taxonomy are defined in terms of entities the ontology says do not exist. I assume this is not the intent.

2. Continuant parthood (obo:BFO_0000176). Then the mereology misbehaves. The reading is admissible -- in the BFO 2020 CLIF, continuant-part-of is typed over all continuants [bdd-1] -- but it was never provided for: continuant-mereology.cl does not mention generically dependent continuants, and its reflexivity [mcd-1] and antisymmetry [tab-1] axioms cover independent continuants only. What does apply globally is weak supplementation [fyf-1] and the unique product [gzr-1], and contents are unkind to both. The content that David replies to Alan and the content that Alan replies to David would be wholes with the same proposition-parts about David, Alan, and replying; likewise that p and q and that p or q. Material cases of same parts, different wholes are no comfort here. The same bricks build a train on Tuesday and a spaceship on Wednesday, and no axiom objects: continuant-part-of is time-indexed [bdd-1], the train and the spaceship never exist at once, and what distinguishes them is the spatial arrangement of the bricks -- a fact about where the bricks are, which is not a mereological fact at all. Both contents, by contrast, exist at once, and a content offers no left or right in which its alleged parts could be arranged; the arrangements actually in the vicinity (word order, glyph position) belong to the bearers and vary freely across bearers of one content -- 'Alan is replied to by David' reverses the order and keeps the content. So: read content-parthood extensionally, in the spirit of the axioms, and the two contents collapse into one ICE, which would be news to both David and Alan (David, as in David L. as in David Lutz); they differ in aboutness and in truth conditions. Read it non-extensionally, and the two contents are distinct without anything making them so -- unless one posits a content-level analogue of the bricks' configuration, which appears nowhere in BFO or CCO, at which point the arranging rather than the parthood is doing all the individuating work. I thought I made this up and then realized I was taught it long ago (when David L. was young) and had forgotten: Lewis pressed it against Armstrong's structural universals -- same constituents, differently arranged, while mereological composition is unique (Lewis, 'Against Structural Universals', AJP 64, 1986) -- and Armstrong's defence was to purchase a second, non-mereological mode of composition (Armstrong, 'In defence of structural universals', AJP 64, 1986). BFO has nothing to help with this.

3. An undefined primitive. Then 'consists of' names a structural relation that appears in neither BFO's nor CCO's relation inventory, and three core branches of the ICE taxonomy rest on a relation that officially does not exist.

There is a reading on which nothing goes wrong, and it costs no new metaphysics. ICE identity is carrier- and encoding-independent -- the same report survives translation, reformatting, and re-serialization, which is what the "actual or potential copies" clause is for. All syntactic articulation (glyphs, ordering, layout, file structure) lives bearer-side, in the patterns of Information Bearing Entities, and CCO already models it there correctly: List assigns its ordering to Information Bearing Artifacts and Sequence Position Ordinality, not to contents. On this reading an ICE has no internal set-theoretic or mereological anatomy, and part-talk about contents has two innocent readings -- literal parts of bearers, or distinct ICEs related by representational involvement (the abstract's content to the article's content) -- neither of which puts pieces inside a content.

Suggested fixes, in ascending order of effort: (i) reword the four definitions to drop 'consists of a set/collection of' -- e.g., "An Information Content Entity that describes some Entity by means of one or more propositions" keeps the plurality in the manner of describing without composing the content out of it; (ii) add a scope note on Information Content Entity recording that ICE identity is independent of carrier and encoding and that part-talk is to be read bearer-side; (iii) if a content-level involvement relation is wanted for document/ontology decomposition, introduce it as a dedicated object property rather than overloading parthood. I am happy to draft a PR for (i) and (ii), if I might be trusted with such a responsibility.

Pointers to any prior thread I have missed are welcome.