On 15/08/25 at 19:02 +0900, Simon Richter wrote: > Hi, > > On 8/15/25 6:26 PM, Lucas Nussbaum wrote: > > > All this is reasonable and meets real needs. As you care about better > > integration of Salsa in Debian workflows, I think that you should review > > what is the current integration level, and how it can be improved. > > There is another debate below this: if Debian's workflow is different from a > standard forge workflow, does Debian need to change, or the forge?
Could it be that the standard forge workflow is optimized for organizations with a few projects, while in many cases, Debian's organizations (teams) deal with many projects (packages) that require an additional layer of consistency that is difficult to standardize? This almost sounds like a different version of pet vs cattle. As long as the count of projects you routinely deal with makes it possible to manage them like pets, then the standard GitLab/GitHub interfaces are fine. But once you reach the cattle level, you need more tooling (somehow built on top of the forge's APIs and on tools like mr) to efficiently manage your collection of projects. And as long as someone stays within a context where pet-like management is enough, it's difficult to grasp why it's not enough for Debian. Alternatively, we could investigate whether it would make sense to merge all Debian packages in a single "project" (either globally or one project per team). That's what the Haskell Team is doing (https://salsa.debian.org/haskell-team/DHG_packages/) and it would be interesting to understand the pros and cons of that approach. Lucas
