Answer A-488 and A-488a: a step ask names its build seat's language; a module following the seat list consumes its events

Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
This commit is contained in:
2026-11-19 13:40:10 +01:00
parent 96be83fcfc
commit 25df48ad70
@@ -2011,3 +2011,13 @@ the principle that settles it, cited as "design 47 G A-n".
names "the newest" (the kept trunk's start read and its read on `catalog-revision`, `TrunkView`, `BranchView`); a
read of a revision by its number keeps failing loudly when that revision is gone. (From the review of the fix of the
live race on 2026-11-19 11:55: trunk.488 removed while the catalog half read its entries.)
- **A-488** (principles 7 and 3) A pipeline step's ask names the language of the build seat it is asked of
(`node-build-go` go, `node-build-node` node, `node-build-image` image), never the artifact's own language word: the seat
decides what it takes, and one place says it. A test holds every build seat's ask language to the seat. (Live
2026-11-19: the pipeline runner asked `node-build-node` with language "typescript", which the seat refuses; no
TypeScript step ran since 2026-11-17 15:40.)
- **A-488a** (principle 3) A module whose runtime follows the seat list consumes the events that reader follows
(`mesh-catalog.seat-definition-changed`, `mesh-catalog.seat-definition-removed`), so its consumer exists and the
reader starts; `mesh-build-node` declares them. (Live 2026-11-19: `mesh-build-node` consumed nothing, its consumer was
never made, it could not judge a named asker and handed every TypeScript ask back.) The catalog check that refuses a
module following the seat list without them is noted, not built here.