My thoughts on repotag

[Michael Schwendt]

On Mon, 19 Mar 2007 15:42:58 +0100 (CET), Dag Wieers wrote:

> > > I'm sure you could have come up with this:
> > > 
> > > 	foo-1.0.0-2.el4.1.fdr < foo-1.0.0-2.el5.0.fdr
> > > 
> > > Which is exactly how it used to be without the repotag. I can only see bad 
> > > intent in the form of your examples.
> > 
> > Do you realise that now you've gone as far as comparing a release digit
> > with a repo tag?
> 
> No, I am not. In fact I have the same number of components.
> 
> 	2	vs	2
> 	el	vs	el
> 	4.1	vs	5.0
> 	fdr	vs	fdr

?? Has the proposal been changed? It used to be repotag at the right, so
that is _three_ fields instead of four. You would compare packages with
three fields against packages with four fields. Unless you want to enforce
a very-specific format for %release to be used by all packagers and all
repos. E.g. an ugly and inflexible 1%{?dist}.0%{?repo}

> > 2.el4.1 wins over all 2.el4.somerepotag
> 
> I fail to see how that is a problem ?
> 
> Since 2 packages from different repositories have no relation.

They don't have any relation? How comes?

> There is no reason why this behaviour is right or wrong.
> 
> In fact I would expect 2.el4.1 to win over 2.el4.repotag even in a single 
> repository context. (eg. when you add a repotag but older packages 
> have not) That's how RPM was designed, that's how I expect it to be.

Sure, except that using %dist is confusing enough for several packagers
already, as they do things like  2.1.fc6 > 2.fc7

With the repotag the situation doesn't improve. It's another string
that enters RPM version comparison.