How kernel development goes wrong and why you should be a part of it anyway

[Salemboot]

Intel was allowed to make massive changes graphics stack: regressions.

The changes made to Reiser3 to remove the Big Lock hurt performance.

Con's scheduler used round-robin for allotment of resources to processes. CFS caused major regressions compared to O(1). Hence why Google removes Ingo's code in favor of O(1).

Ext4 has huge regressions with various tests run by Phoronix.

In the real world, stable stays untouched, feature complete. Company can't afford breaks. Takes Millions to fix.

[Garegin]

@Salemboot your talking about the kernel, propably the most stable piece of software in the linux stack. the rest is mostly alpha quality cack compared to windows, osx or other nixes. if finder was as buggy as nautilus there would have been mobs of people with firetorches.

[moveaxebx]

@Salemboot

Reiser is dead.

Con's scheduler works good only on desktops.

Ext4 shouldn't be considered FS for production environment...at least until they introduce "bigger" block size and other features.

[shaurz]

I though btrfs was dead? I'm glad to hear it's still being worked on.

[moveaxebx]

@shaurz

Dead? Far from that. Check out btrfs mailing lists.

[shaurz]

@moveaxebx It was just a rumour I heard on reddit. I am hoping to switch from ext4 to btrfs in the future when it is mature.