diff --git a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md index 407f0365..87c149a1 100644 --- a/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md @@ -711,3 +711,9 @@ The lead's one-line answers to the questions H2's develop and review raise (impl tool runner's own tests of the objects delete, where it already is or where it is next touched (noted for H2b if it is not there); no new whole test of distribution-tools, and mesh-tools, approved tonight, is not reopened for it (rules-are-weighed, said here). H-A-453 (3)'s RG3 line is amended so. P3. +- **H-A-454 (second addendum, 03:30)** Corrects the first addendum, which contradicted ADR 0436 §6 (a work-queue stream + is read by exactly one process, named in its `bus.queues`; the validator holds it). The builder's process gains a `bus` + of its own in the manifest, naming `publishes` in `queues` and everything else it does on the bus (emits, consumes, + invokes, its state, asks), so it runs on an account of its own granted exactly that, as the runner's process does; the + validator's rule stands unchanged. A grant the process lacks at run time is a loud fault of the declaration, caught + by the module's whole test. ADR 0436 §6, P-failure-loud.