Design 47 H: H-A-329, a validator run by hand reads the seat list the mesh holds

This commit is contained in:
2026-11-21 21:30:49 +01:00
parent 5367dc409f
commit e828d8a61f
@@ -1982,3 +1982,11 @@ Where two preparers disagreed, the chosen one is named.
the answer is published; the match is by the package's build fingerprint; the publish event stays the second
path. The seat mesh-builds is defined in the delivery repository's own contracts module, so the whole change is
one branch there (the lead's map named the contracts repository wrongly). P-NATS-native (a call where an answer is needed; the owner answers), P7, ADR 0596.
- **H-A-329 (a validator run by hand reads the seat list the mesh holds, 21:30)** Met twice tonight: a validator run
against the repositories' trunk seat files refused the builder's and the runner's manifests on fifteen and three
lines (registry seats' verbs the validator program does not see in those files, which declare acts and accepts) that the live
validate of the same modules passed today; every developer and reviewer runs the validator this way because the
live seat list's answer is too large to read into a session. Owed, H2: a tool of the seat mesh-catalog answers the
seat list as one object a validator run can take (paged or by digest), and the validator program takes it by address;
and a validate step that touches a module counts only the problems new against its before case (causedProblems), as
the reader found it does not. P-failure-loud, one-question-one-tool.