Ankur Sinha via python-devel venit, vidit, dixit 2025-06-25 14:19:28: ...
So, we're thinking of:
- prioritising packaging/maintaining software that is unavailable on PyPi (or difficult to install from there)
- dropping software that is pip installable from Fedora
- shifting focus to testing that the software we care about is pip installable on current Fedora releases
- documenting software status (pip installable or not) and any additional tips/steps required to get it working and informing upstream of issues that we encounter and submitting fixes where possible
...
What do you all think about this? Please share any thoughts you have about this, and any tips/tricks/tools that can be used. If it's something we can do for Python software in Fedora in general, we'd be happy to do that too.
Thanks for this proposal. I think it's coming just at the right time and should - as you indicate - help us focus on Fedora as a platform. In particular:
- The separation between "platform" and "application" wil become more and more important when the platform is provided as an immutable OS (for more users). Neither layering nor rolling your own (bootc) image make much sense for an "app", nor (in many cases) a flatpak.
- Reducing our set of rpm-packaged python apps helps whenever there is a major py change, be it a mass rebuild for a new python version or a change in packaging (such as dismissal of `setup.py install`).
I like the idea of having a "Fedora vetted/tested subset of PyPi", i.e. packages confirmed to be installable on Fedora X for all X in current releases, a bit like koschei for PyPi, possibly using existing `%check` or such. More ecosystems than just Python could benefit from this.
Michael