On Fri, 2005-05-27 at 10:14 -0400, Matthew Miller wrote: > On Fri, May 27, 2005 at 03:43:23PM +0200, Ralf Corsepius wrote: > > > knowledge of things that need to be done (e.g. on FC4 it will remove the > > > i386 version of perl on x86_64 installs), information that isn't > > > available to yum or apt. > > Well, that's what I call broked design. > > Maybe. It's easy to call things broken in retrospect. However, much of the > "magic" that anaconda does is outside of the scope of individual packages. > It drags along knowledge like "hey, if you had mozilla installed before, you > probably want firefox installed now, even though the two packages can > coexist just file", This is a different class of problems: It is adding packages, not replacing packages. It won't harm users, nor will it have ugly side effects. > and also "hey, lvm1 needs to be converted to lvm2". This is a similar class problems to forcing an "arch switch", however if anaconda can handle this issue, rpm probably can do it, too (By using rpm-tag tricks in combination with %pre/post etc. scripts). However, I would not do this inside of an rpm and leave it to the user in this case, until he explicitly requests such switch. Nevertheless, an "arch-switch" of FC packages probably is more harmless than that. As I said in my reply to Rahul, I know too little about the internals of this particular problem, but I doubt this can't be handled by incrementing E:V.R. Ralf
