Reading back my message I realise I may have been a little harsh.
IMHO you take too much as a given the fact that the current process can
be extended to many comps files covering the full FC+FE perimeter.
So far FC has mostly lived on a comps file legacy that took many
releases to write. With vastly simpler needs, FE didn't get a satisfying
comps (satisfying as pretty complete or even syntactically correct) for
a several releases. So I seriously discuss the "simplicity" of writing a
Fedora-level comps from scratch. We haven't for quite a long time and
when we did we sucked at it.
In a bigger, more dynamic repo, with multiple parallel release sets I
just don't see every user getting the comps it needs through "simple"
comps duplication. Some posters have written repoview could just source
another metadata file for classification. I didn't see them volunteer to
write or maintain this file by themselves.
Now of course if everyone thinks piling up comps files is no problem,
and that each new comps user can just start from an uncategorized
package list, I'll shut up now. I'll just offer the following
Why should I, as a FE contributor, maintain my package entries in any
comps file? If adding a new comps is trivial, any comps is no less
trivial than the others. If uncategorized packages are not a problem for
other comps writers, surely they aren't more a problem for the anaconda
comps team either? Why shouldn't I just push my packages without caring
one bit about compsing? RELENG can start from raw package lists like
everyone else, if it's so easy.
fedora-extras-list mailing list