On Mon, 2007-09-24 at 09:54 -0500, Chris Adams wrote: > Why keep the i386 versions in the repo? It would be better to write > down the rules that are used to pull i386 packages into the x86_64 > repo and then make a yum plugin that can point to the i386 repo and > follow the rules (when enabled). There's a lot of sense in this. One of the options being discussed is to ship _nothing_ of the secondary arch (or almost nothing) in the primary repository, and allow yum to be pointed at the basic i386 repo. Of course, that means we've suddenly got a whole load of i386 packages visible to x86_64 yum which weren't there before; with the current install-both behaviour that wouldn't work too well. But as part of a more coherent approach, I think it's the way forward. Note that for some 64-bit architectures like sparc64/ppc64 we run mostly 32-bit userspace, because 64-bit userspace just isn't worth the bloat (the only reason it's worth the bloat for x86 is because you actually get a sane number of _registers_ that way). So the 32-bit tree would probably still carry a _few_ 64-bit packages there (kernel, gdb, systemtap, etc.). -- dwmw2 -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
