Michel D�nzer wrote: > Adam C Powell IV wrote: > > > On a "philosophical" note, since they are so close to working, I'd prefer to > > use stock kernels and bug report against those, with as little patching as > > possible, it seems this will hasten the day when unpatched kernels will be > > usable across more platforms. > > I wouldn't count on this ever happening. You might want to search an archive > of linux-kernel or Kernel Traffic (http://kt.linuxcare.com) on the topic.
Hmm, this is philosophical and drifting off-topic... But if people are really resigned to the situation as you describe it, then it seems pointless to put other arches into the main kernel tree at all. Why not just completely leave them out and have separate people maintain separate trees for the non-i386 arches? When I used to be exclusively alpha way back in late 2.1 days (okay, so it's not *that* long ago, but it's when I started using GNU/Linux :-), about 1/3 of the kernels wouldn't build, and half of those that built wouldn't boot, but we stuck it out and didn't maintain a separate tree and patched the pre-releases, and stock 2.2 had seamless alpha support as a result. Thanks to people like Richard Henderson and Andrea Archangeli, Alpha seems to always be working or close to working, with active discussion of and patches against the *stock* tree appearing on debian-alpha, [EMAIL PROTECTED] and (to a much lesser extent) [EMAIL PROTECTED] This seems an intelligent way to maintain an architecture on a single kernel tree. Having five people maintaining five PPC trees and occasionally sending patches to Linus seems the wrong way to maintain an architecture, and an implicit resignation to the total worthlessness of any PPC material in the stock tree at all! From what I've heard, sparc(64) support in stock is pretty good, m68k too, why shouldn't we expect the same for PPC and ARM? (Oh- and stock 2.2 works decently on PPC too- I haven't had any problems). I understand (from my reading of K-T) that features will appear in new kernels for i386 which break other arches, but then we should *try* to use the new features and patch against the main tree, right? Maybe this is related to Eric Raymond's "curse of the gifted" post, and we really need to use more sophisticated tools for the kernel, like a tinderbox server of some kind... Just wondering, -Adam P.
