251
Gentoo Linux Begins Codeberg Migration In Moving Away From GitHub, Avoiding Copilot
(www.phoronix.com)
A community for everything relating to the GNU/Linux operating system (except the memes!)
Also, check out:
Original icon base courtesy of lewing@isc.tamu.edu and The GIMP
Moving nixpkgs development from GitHub would be ambitious, as that repo already pushes their infrastructure limits with enterprise level support.
https://discourse.nixos.org/t/nixpkgs-core-team-update-2026-01-22/74585
I suspect the Nix org would need to garner many more sponsors to fund the hosting expenditures for an equivalent forge with matching CI/CD, PR automation, and geo redundancy. Would be nice to see.
I fully agree that nixos moving off GitHub is a nonstarter for technical reasons. The nixpkgs repo is enormous and has caused loads of infrastructure problems at GitHub over the years.
But the NixOS foundation currently has over 300K Euros in their open collective account. They get free sponsorship from GitHub for the repo and free sponsorships covering the very expensive build and caching services too.
https://opencollective.com/nixos
@patrick @ruffsl Don't lose sight of the fact that Github is actually Microslop, and they'll do anything to keep their fingers in that pie.
It is quite a large project to move from over ci/cd to another. I worked at EA for 13 years and we moved from Jenkins to another platform. (Jenkins suuuuucks. If you ever wonder what hell looks like, try to maintain a Jenkins instance with a shit ton of plugins, half of which are no longer maintained and you have to upgrade the instance...)
How does it push the limits of enterprise level support?
And I don't think they'd need sponsors but volunteers to host the CI machines. Tapping into the community to provide those could possibly get a few. If effort were put into using some content addressed network, distribution of packages could further be spread.
And if radicle finally supported large repositories, it could actually be hosted there and distribution would happen across all nodes willing to host the project. But that's not possible yet.
A lot of effort is being put into building a ecosystem for a proprietary Microslop platform. It's a pity.
This discourse post has more details. Essentially: Nixpkgs is an 83 GiB repository with 20,000 forks that incurs a ton of API usage from CI and bots. It turns out this caused some scaling issues. It's unclear to me if other forges would have run into exactly the same issues but if the repo can cause replication failures at GitHub it can almost certainly cause problem at Codeberg.
Edit: For comparison, the Gentoo repository is 20 GiB and has 2.2k forks (on GitHub). The largest repository on Codeberg is 30 GiB, but that's just a blocklist data mirror with no CI.
Edit 2: Also, since I'm reading through Gentoo's mail archives for this they also seem to be at least slightly worried:
The 83GB figure is the entre nixpkgs network not nixpkgs itself. Before codeberg got that far, it would take a while. It took nixpkgs 20 years to get there. The actual nixpkgs repo is 2.5GB bare and 5GB checked out. Github could still function as a read-only mirror (as in no PRs, but it's pushed to from codeberg or whatever forge they use).
These aren't unsolvable problems. It's also a reason I hope radicle will get contributions to support bigger repos as it truly is a distributed forge.