%{?dist}, recommended or optional?

[Jeff Spaleta]

On 1/8/06, Roozbeh Pournader <roozbeh at farsiweb.info> wrote:
> So, should we try and make a list of situation where %{?dist} is not
> really needed (so we can recommend it in other cases)? There is already
> one candidate: a font package in which fonts.cache-2 is ghosted in the
> FC3/4 package.

I think this is a waste of effort to attempt to make an exhaustive
list of situations where ?dist is strictly needed or strictly not
needed... simply because its never strictly needed. It is a tool which
a maintainer may choose to use. It is documented in the wiki because
it is the ONLY allowed mechanism by which distribution revision can be
encoded into the release field of the packages.

If anyone is confused as to the recommendation in the wiki for a
specific package, they can come into this list and get more opinions
as to what to do for the specific case.  Unless the steering committee
is prepared to make use of the ?dist something reviewers are suppose
to check for... this level of detailed documentation as to why and why
not you would choose to use it seems out of place to me.

This simply isn't that important.  ?dist doesn't automatically
garuntee a workable upgrade path. Maintainers need to be educated
about how the ENVR comparisons work in rpm instead of focusing on the
list of reasons to use or not use ?dist.  If maintainers want to mark
the distribution revision in a package.. then they must use ?dist. If
maintainers don't want to mark the distribution revision in a package
then they don't have to use ?dist at all.

If anyone else other than Roozbeh here is confused by the
recommendation... then I suggest that all we need to do is ask spec
authors to drop a comment in their spec saying why they didn't use
?dist as recommended in the wiki.  That way anyone in the future can
examine individual specfiles to learn what individual maintainers
thought.

-jef"damn the semantics! I'm going in!"spaleta