I’m a supporter of rolling release mode for kernel. A new kernel means better hardware support and it has helped me a lot several times. Unfortunately, pushing new kernels into a stable release has its evil side – regressions. I usually have no problems with regressions when a new kernel comes. Unfortunately, the kernel 3.5 has brought a lot of regressions and problems:
- After several hours of uptime, I experience artefacts, disappearing items in menus etc. (#848099). In my case, it occurs on two different Intel cards and AFAIK I’m not the only one affected by this regression.
- Many users have reported severe problems with ATI cards (#846505). I’ve heard there are quite a lot of regressions in nouveau, too. My friend couldn’t suspend/hibernate his laptop with an nVidia card after upgrading to 3.5.
- NFS problem that affects for example oVirt (#845660). oVirt developers even recommend running Fedora with pre-3.5 kernels.
IMHO 3.5 is definitely the worst kernel since Fedora switched to the rolling release mode for kernel. It has fully showed the drawbacks of this model.
BTW all the problems mentioned above are solvable by downgrading to 3.4 (you still can find them in Koji).
How can we make it better? Should we be not so aggressive in pushing new kernels into stable releases, doing more testing? Or should we give users an easy way to stick with an old kernel or downgrade to it? (of course, users have 3 kernels to choose in GRUB, but sooner or later the older version gets removed and if users upgrade or install Fedora after the new kernel is released they don’t have the old kernel at all).