[Yum] yum chooses 'older' provides

[seth vidal]

On Fri, 2006-12-08 at 23:21 +0100, Patrice Dumas wrote:
> On Fri, Dec 08, 2006 at 12:28:08PM -0500, seth vidal wrote:
> > 
> > comparing version numbers of 2 packages that don't have the same name
> > doesn't do any good - you're comparing apples to oranges and since
> > version numbers are arbitrary w/i any application you have no idea what
> > you'd be getting. At least the way it is it's deterministic.
> 
> Comparing 2 packages which provides a versioned provides with the same name
> is not comparing apple and oranges. If I have 2 packages who have
> 
> Provides:  foo = x.y-z
> 
> and 
> 
> Provides:  foo = A.b-c
> 
> comparing x.y-z and A.b-c is comparing packages with virtual provides 
> with the same name. Using the version of the versioned virtual 
> provides may not be always right, but not more nor less than using 
> the shortest name. Agreed, there may be 2 packages which changed name 
> and still have a virtual provides with the old name, and in that case the
> result is not right. However, I can't think about a case where using 
> the virtual provides version is worse than using the shortest name, 
> while there may be cases (like the one I presented above or the libnet)
> in which it is the right thing to do.

So, in short, there is no right thing to do b/c all of the issues run
into another corner case. The correct behavior is to use packaging
policy to remove packaging problems.

-sv