From 260e2c0efcd5dae5891557967e14cc7e996ec52b Mon Sep 17 00:00:00 2001 From: jochen Date: Sat, 21 Nov 2026 11:00:39 +0100 Subject: [PATCH] Design 47 H: H-A-235, a step ask nobody answers within its bound is asked again, then ends could-not-run --- 03-DESIGN/01-to-be/47-milestones/H-lead-answers.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md index 95dd6f27..9f2c97aa 100644 --- a/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md +++ b/03-DESIGN/01-to-be/47-milestones/H-lead-answers.md @@ -1312,3 +1312,10 @@ Where two preparers disagreed, the chosen one is named. three-repository group for words alone; the holder's not-found answer and the indexer's fallback carry the behaviour, and the comment follows with the next contract change (owed, H2). Said as a weighed bend of rules-are-weighed. P7, P4. +- **H-A-235 (a step ask nobody answers within its bound is asked again, then ends could-not-run, 11:00)** Met live: + a pipeline's validate ask was lost when the seat mesh-wants' holder was restarted, and the runner waited on it for + nine hours with its one slot, every later pipeline behind it. Every ask the runner makes to a seat (a build, a step, + a validate, a composition) has a bound from the seat's contract or the runner's setting (ten minutes without the + holder's first word); past it the runner asks once more, naming the ask asked again, and past a second bound ends + the step could-not-run with "its ask was not answered" and a finding, so the pipeline ends and the slot frees. + P4 (failure must be loud; nothing waits for ever), the instruction nothing-moved-in-five-minutes.