On Thu, Aug 7, 2025 at 4:30 AM Andreas Enge <andr...@enge.fr> wrote: > [...] > sorry for not being clear. The change would be put on the "world-rebuild" > branch, which would be handled like any other branch and built by QA. > But since there is no competent team behind it, every commit should be > handled mainly by its author when problems occur. This concerns > repairing resulting breakage, for instance, or suggesting to drop the > commit because eventually it does not work as intended. So in a sense, > the "team" behind such a mixed bag of commits could be their authors.
There are two cases where I think Guix should invert the obligation in order to scale. Direct and obvious problems can be handled by the author, but for complex issues it would be good to have package "advocates" with the necessary experience and knowledge. This would preempt the current state where packages are broken post-rebase. Even just receiving a notification that a subscribed package is broken on master or QA. This would also shift the onus for updating dependent packages, rather than obligating the author, which has led to many packages remaining years out of date. And if no one advocates for the package it can be removed. The second case is package archival. Rather than store infinite builds indefinitely, make it simple for users to backup the artifacts necessary for a reproducible build and aggressively garbage collect the servers. Greg
