You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
OQ-2 — Do seller catalogs (catalog_ref) belong in 3.0? #3
Or should the object stay fully inline until a later version? catalog_ref
buys compactness across a large shelf at the cost of a resolution step during
discovery and a versioning regime sellers must operate.
What would settle it: how large a real shelf gets, and whether sellers
already maintain versioned, immutable catalog entries or would have to build
that.
This is the tracking issue for open question 2. Answer it here, in the comments. Evidence from implementation or real inventory carries the most weight (comment process).
Comment period: 2026-09-22 → 2026-10-22. This repository is public, so anonymise any examples.
One security consideration for OQ-2. A catalog_ref resolves to seller-controlled content that buying agents fetch at discovery time and, for prose entries, feed to a model. If catalogs stay in 3.0, two things make them safe to dereference at machine speed: a content digest so the pinned version can be checked (#129), and the same rules as for inline free text, so typed fields govern and resolved text never directs the reading agent (proposed in #155). With both, catalogs cost no more trust than inline values.
Activity
pkanna000 commentedon Nov 7, 2026
One security consideration for OQ-2. A
catalog_refresolves to seller-controlled content that buying agents fetch at discovery time and, for prose entries, feed to a model. If catalogs stay in 3.0, two things make them safe to dereference at machine speed: a content digest so the pinned version can be checked (#129), and the same rules as for inline free text, so typed fields govern and resolved text never directs the reading agent (proposed in #155). With both, catalogs cost no more trust than inline values.