Personality can be a huge problem because the kernel and its parts are a long term project and I would prefer people that can work together for its entire duration.
Accusations like "/me sees Ingo run off to find the right combination of hardware and benchmark to prove his point." is far from the level of civility I would entrust my favorite OS's kernel. The proper response would be something in the line "it will suck on this benchmarks, but it will excell in these others and in these hardware configurations" and show numbers. It's always possible to act in a civilized way.
That said, CK seems only a bit off when compared to a couple others. There are a lot of people who like to be quite caustic from time to time. I must confess I have to refrain myself frequently and HN is not LKML.
And, more important: this whole discussion is more or less pointless. It doesn't matter how good a modern scheduler is, all of them impose a very little penalty on a modern kernel. Are we really going to fight this bitterly over a couple percentage points?
Would that be so hard as to include a "-scheduler=" switch in the kernel for it to be selectable at boot time (and so I can pick one that fits both my netbook and my SGI 4096 CPU monster when I want to)? This would pretty much end this discussion.
Accusations like "/me sees Ingo run off to find the right combination of hardware and benchmark to prove his point." is far from the level of civility I would entrust my favorite OS's kernel. The proper response would be something in the line "it will suck on this benchmarks, but it will excell in these others and in these hardware configurations" and show numbers. It's always possible to act in a civilized way.
That said, CK seems only a bit off when compared to a couple others. There are a lot of people who like to be quite caustic from time to time. I must confess I have to refrain myself frequently and HN is not LKML.
And, more important: this whole discussion is more or less pointless. It doesn't matter how good a modern scheduler is, all of them impose a very little penalty on a modern kernel. Are we really going to fight this bitterly over a couple percentage points?
Would that be so hard as to include a "-scheduler=" switch in the kernel for it to be selectable at boot time (and so I can pick one that fits both my netbook and my SGI 4096 CPU monster when I want to)? This would pretty much end this discussion.