Design 47 H2: H-A-454, publishes in the old section, read by a seat verb

This commit is contained in:
2026-11-22 02:25:19 +01:00
parent 93039639e5
commit 5aa3b4b8c4
@@ -684,3 +684,12 @@ The lead's one-line answers to the questions H2's develop and review raise (impl
whose count stands where it was is accepted this milestone (the fold of nine modules' tests would cost more than the
rule protects tonight: rules-are-weighed, said here). The fills gitea's caps rest on are written in a commit (H-A-437),
and tweetnacl is declared where test/fakebus.ts imports it. P3, rules-are-weighed.
- **H-A-454 (publishes stays in the old streams section; its reads are a seat verb, 02:58)** B4, B13; corrects
H-A-448's first half, which contradicted ADR 0619 §1 and §820 (a part of the foundation keeps its bus state in the old
sections, new entries included; the validator holds it). (a): `publishes` is declared in the builder's old streams
section as H-A-446 (2) said (retention work-queue, duplicates 24 h, max-deliveries 10, the process's bus.queues), the
consumers made per holder at run time as H-A-440 (1); and its reads take builder_health's form (H-A-435 addendum): the
seat mesh-builds gains the verb `publish-queue` (a D row, the builder's developer), served by the builder's process from
its own stream (the counts per holder and class, the oldest and newest, a capped list), and the tools program's
builder_publish_queue calls it, as the runner's tools call its seat's verbs. builder_publish_drop and _first stay as
H-A-448 has them (reasks the builder performs). The validator gains no exception. P-NATS-native, P3, ADR 0619 §1.