Jason L Tibbitts III
tibbs at math.uh.edu
Sat Mar 28 01:07:47 UTC 2015
>>>>> "MM" == Matthew Miller <mattdm at fedoraproject.org> writes:
MM> Basically, this is an end-run around the requirement of doing
MM> individual package reviews for a zillion completely separate
MM> packages, right?
That was my opinion, but you could argue the same for Perl, I suppose.
We're essentially packaging a complete distribution. There aren't too
many examples of that around.
My proposal was to machine-generate the individual specs and have FESCo
grant an exception to have them reviewed in a block. The hardest part,
of course, would have been the licensing, except that texlive had
undergone a rather complete license audit and every single file has been
cleared. I don't know if that's still valid.
MM> Since this approach really has disproportionately large negative
MM> impact on the rest of the distro, it seems like we should find a
MM> better way.
Well, the "better way" (in my opinion) would involve:
Making sure the licensing situation is still good.
Getting CTAN hooked up to Antiya. Mechanically add every single texlive
package that has an acceptable license. (I don't know if texlive
actually carries any that don't.)
Mechanically generate a huge load of srpms.
Get FESCo approval for a variance to the usual package review procedure.
Have three or four experienced reviewers go over a sampling of the
packages and the generator. (which I imagine would only be used once.)
Mass import them all into rawhide.
This all needs to happen soon or wait until F23 branches. I think it's
quite doable, but will need some scripting and coordination with
infrastructure for some of it. And of course the question arises
whether some other process will break.
Now, there's a question as to how this all interacts with the texlive
model of doing rollup releases, and whether that would cause any
problems with texlive upstream.
More information about the devel