On Sat, Dec 5, 2009 at 10:24 AM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > You seem to be under the incorrect assumption that we primarily want a fair > scheduler. Our main interest it to get a scheduler that "works well" in Haiku > (yes, totally fuzzy specification). If a fair scheduler does achieve that -- > great! I wouldn't ignore other algorithms, just because they aren't fair. I think in general just creating something that works well is a good idea, regardless of what buzzwords it has or not. But it seems the fair schedulers have worked well for Linux so it probably doesn't hurt to try that too. But it might be smart to try that later after other approaches have proved less than ideal. I suppose it could be summarized that trying to create a perfectly fair scheduler to start off with is a bit like premature optimization. -- Regards, Ryan