From 28def7eae538524244e074a4e38a005a18a7595f Mon Sep 17 00:00:00 2001 From: jochen Date: Sun, 22 Nov 2026 00:17:01 +0100 Subject: [PATCH] Design 47 H2: H-A-393, a version held at another digest is refused before the reservation is asked --- 03-DESIGN/01-to-be/47-milestones/H2-lead-answers.md | 8 ++++++++ 1 file changed, 8 insertions(+) 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 d501685f..8efdc5fa 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 @@ -184,3 +184,11 @@ The lead's one-line answers to the questions H2's develop and review raise (impl line per directory set right, so a drift lasts at most until the engine's next start. An unchanged composition is still not applied (idempotence keeps). Test: a record-held directory with another mode is set right at the engine's start; one the mesh never made is left and said. P-mesh-owns-the-truth, issue 339. +- **H-A-393 (a version the store already holds at another digest is refused before the reservation is asked, 00:44)** + Addendum to H-A-387. While the reserved digest is still empty, a second node's publish of the same version with + another digest would pass Recorded, and the OCI store's Put would overwrite the tag the first publish set; the gitea + forge refuses this itself, the artifact registry's store does not. A-490j holds for every holder: "a second publish of + the same name and version with another digest is refused as the forge holds the first". So the distribution module's + Publish refuses, in its Held branch before Recorded is asked, a publish whose version the store already holds at + another digest, naming both digests; the same digest is taken as today. Test: a table row (version held at digest bb, + publish of aa, refused naming both). Same fix branch. P-failure-loud, A-490j.