repository breakage

[Alex Lancaster]

>>>>> "sv" == seth vidal  writes:

>> https://devel.linux.duke.edu/bugzilla/show_bug.cgi?id=296
>> 
>> It seems that given that the presence of any third party repos that
>> can block FC+FE security updates (and even FC updates can sometimes
>> be blocked by FE packages that haven't been upgraded!), wouldn't a
>> prudent approach be to allow at least the packages that aren't
>> blocked be upgraded and hold back the other ones (as apt-get does)?

sv> We can't KNOW what the value of one package is over the other. We don't
sv> have any severity information in the metadata. How can we rank
sv> which package is a more important update?

Well, if we don't adapt the tools to cope with out-of-sync packages,
then we need to shift the onus to the packagers and ensure that
updates are shunted out that are compatible with (at least) the
official downstream repos (e.g. FC updates should be compatible with
FE updates).

I've lost count of the number of bugzilla's I've filed against galeon
for rebuilding when a new version mozilla came out, because it caused
yum to grind to halt.  It's that kind of breakage that makes people
think (fairly or unfairly) that Extras is kind of second-class to
Core, because it is very rare that a set of updates from Core ever
cause yum to not upgrade.  It seems that the Core maintainers aren't
necessarily following Extras to see what Extras packages would be
impacted.

It seems the problem is more a matter of co-ordination: Core updates
come out and Extras sometimes have to scramble to keep up, then Extras
updates come out and Livna plays catch-up.  Perhaps there is better
way of using repoquery and/or other tools by ensuring that at any
given time the combination of at least FC+FE (or even better
FC+FE+Livna) is consistent before updates hit the mirrors.

Alex