[CentOS] Selectively updating protected repos

Mon Jul 3 20:31:50 UTC 2006
Matt Hyclak <hyclak at math.ohiou.edu>

On Mon, Jul 03, 2006 at 04:18:44PM -0400, William L. Maltby enlightened us:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> > 
> > On Mon, Jul 03, 2006 at 03:39:55PM -0400, William L. Maltby wrote:
> > > Now if we can be sure that Yum understands that .rf, .plus, ... are all
> > > the same packages, we're home free. But ... are they the same packages?
> > > This may be a dumb question, but can we be sure firefox from one repo is
> > > compatible with what a user runs? Kernel features differences, etc. For
> > > firefox, maybe the answer is yes. Does that apply generally to other
> > > packages too? Is the "unprotect" solution certain to be "generally
> > > applicable"? OTOH, that is a user administration problem, isn't it?
> > 
> > Actually, that is a minor issue. RPM itself has the final vote on
> > who is the newest package. So YUM really should not worry about it.
> > At least, I hope YUM uses rpmVersionCompare.
> 
> You just provided the "connector" in my thinking. Another poster
> mentions DB, RPM has a DB, RPM's DB retains the external name, the
> external name contains ".rf", ".plus", ... Voila! Yum can, if it
> chooses, use the RPM DB, tell which repo provided the currently
> installed package instance, combine that with includepkgs/exclude,
> (un)protect, .... Another DB (yum.conf, repos, ... or a DBMS) can
> provide some additional "rules".
> 

You need to make sure you understand that the "name" does not contain .rf,
.plus or otherwise. That is (generally) part of the Release String. RPMs are
named in a Name-Version-Release.Arch pattern. 

This has been discussed on the yum mailing list previously, and it is
difficult (at best) to track what repository an RPM came from, especially
when RPMs are installed by hand, repositories are removed, etc. The RPM
database does not allow for this information in its current implementation. 

In order to implement what you suggest above goes beyond yum into RPM-land.
Either way, this probably isn't the right mailing list for this discussion.

Matt

-- 
Matt Hyclak
Department of Mathematics 
Department of Social Work
Ohio University
(740) 593-1263