[fab] Non-standard kernels in the Fedora Multiverse

[seth vidal]

On Tue, 2006-05-09 at 00:11 -0400, Jesse Keating wrote:
> On Tue, 2006-05-09 at 00:01 -0400, Greg DeKoenigsberg wrote:
> > Here's the fallback position: Fernando continues to maintain the CCRMA
> > kernel in his own yum repo, and *everything else* gets pulled into
> > Extras
> > over time.  (To the best of my knowledge, none of the CCRMA apps
> > *require*
> > the CCRMA kernel -- it's just a huge help for getting any actual work
> > done.)  That way, at least Fernando has a mechanism to spread the
> > workload
> > for maintaining CCRMA among several assistants, and can spend most of
> > his
> > time maintaining his own kernel as he sees fit.
> > 
> 
> Or do we fire up thoughts on Alternatives again?  Somewhere that we can
> host replacement packages that folks can use to assemble 'Fedora'
> variants but not be tied to the kernel or whatever.  If we use the same
> rules, or come up with a good rule set for Alternatives, same package
> quality, same build systems, etc... we should be able to call it Fedora.
> 

Some complexity in enabling Alternatives:
 1. we can't enable alternatives by default - the obsoletes it could
allow would eat packages for people who really just want to use core.
 2. create a sensible way of dealing with conflicts - something we don't
really need to deal with right now.
 3. dealing with alternative tree creation and QA. What if a user
creates a fedora 'distro' using an alternatives kernel? How does that
impact testing? How do we cope with the near endless number of
combination or configurations we might get?

-sv