diff --git a/03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md b/03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md index 60c9470a..5053f250 100644 --- a/03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md +++ b/03-DESIGN/01-to-be/47-milestones/G-22-lead-answers-2.md @@ -1902,3 +1902,8 @@ the principle that settles it, cited as "design 47 G A-n". needs stays as it is: the backup before the apply (§296) and the operator's yes on a failed backup or a held revert (§710) are the node-engine's and the asker's, not a hold of the want. (Live, 2026-11-17 20:57 and 21:15: feh 0.1.8, released 21:39 with its pipeline passed, was held on shanks and g14 by both holds, though nothing depends on feh.) +- **A-478b** (A-478; principles 7 and 2) The order waits in the step place stay: the backup holder and the bus module + move first, the node-engine last, and at a revert the module taken in first. They order one machine's own apply and + keep what a module holds safe; they never hold a want for another module's delivery or for the build an acting part + runs, which is what A-478 ends. "The assigner's whole want rule" in A-478 is the rule that chooses the version; the + order waits choose only when, on that machine, the want's apply goes. (From the review of A-478.)