A pack relation is a state by construction

[WaylandYang]

Changes one decision record only — the status line of 0031. No code.

#593 made corpus-grown relations take temporal from the proposal. The next cut was going to classify the relations that ontology packs bring in — OWL import writes 'state' for every property, and packs are the cold start (0008), so most relations in a real base come from there. Before writing that cut the five packs were measured:

pack	object properties	events
schema.org	946	0
PROV-O	60	0
ORG	32	0
IOF-core	73	0
FOAF	—	0

The 16 with event-like names (acquiredFrom, publishedBy, releasedEvent, …) are, on reading their comments, all states pointing at an agent or an event node. Standard vocabularies reify an event as a class (PublicationEvent, prov:Generation); their object properties are states by construction. IOF writes the temporal semantics into the property name, and its …AtAllTimes family is a candidate for eternal, not event; it is left alone.

So create_relation_types_bulk writing state is right by construction, not an omission. Event relations come from the corpus (#593) and from a person on the ontology page — and since world_axis looks the predicate's temporal up at read time, a change there takes effect on existing rows at once, with no migration. Recorded so that nobody comes back in six months to "fill in" this cut.

🤖 Generated with Claude Code