A proposal has been submitted for comments that would revamp the way Red Hat's Fedora Linux distribution is developed.
The "Fedora revamp" proposal can be found on the Fedora-devel list
after being published on Monday.
The key changes that the developers behind the proposal seek to achieve is making Rawhide (the latest Fedora development packages) more reliable for installation and use by developers thorugh coherently introducing changes, defining interfaces that applications can rely upon, and ensuring functionality implemented in Fedora doesn't unintentionally regress.
The proposal calls for three levels of stability by having a long-term ABI for applications that won't unnecessarily break, a base design, and an internal API that will hopefully not change within a release life-time. Of course, having a long-term ABI in the Linux world would be a serious challenge with important upstreams like the Linux kernel specifically not being after API/ABI compatibility. Fedora Rawhide is also often quick to pull in unreleased/development packages where API breakage can be frequent and go unreported.
The proposal also calls for more tests to cover stable APIs to fend against accidental breaks and complex changes not affecting other tiers of the stack.
While this fundamentally different Fedora development process may sound at first like some radical user that independently proposes changes, like the crazies out there that have called for Linux abandoning all architectures but ARM
and dropping Ethernet/multi-user/multi-monitor support
, this proposal does have some legs. This proposal was originally discussed in detail in January at the FUDCon Lawrence developer event for Fedora developers. The discussion also continued last month at the Brno Developer Conference at Red Hat in the Czech Republic.
It's possible we may see improvements made to Fedora Rawhide so that it can be more dependable and friendly throughout the Fedora development cycle, but it's unlikely to end with a mandate for a stable Linux API/ABI. Any improvements to the reliability of Rawhide would certainly be welcome by end-users and improving the automated test coverage of important Linux components will benefit everyone.