More efficient package reviewing

[Matthew Palmer]

On Thu, Aug 03, 2006 at 01:48:49AM +0200, Stefan Potyra wrote:
> Am Donnerstag 03 August 2006 00:09 schrieb Lucas Nussbaum:
> > == Do we really want more software in Ubuntu ? ==
> >
> > It seems that the main opinion inside MOTU is "let's get as many
> > packages in as possible". However, most of such packages never make it
> > to Debian. Why ? Maybe because Debian is slower. But also maybe because
> > nobody cares about a 7th package providing the same functionality. In
> > Debian, ITPs are often answered with comments such as "what does it do
> > better than XX ?". Maybe, for new software, we should add a field in
> > REVU where the uploader will describe why he/she thinks the package
> > should be included in Ubuntu, and exactly how it compares to similar
> > packages already in the archive.
> 
> I strongly object: imo FOSS is about choice and thus we shouldn't limit choice 
> just because there is a different package with overlapping functionality. 
> Imho if the person who packages the software cares about it, that should be 
> enough to get it into ubuntu (and of course it is packaged in a sane way).

This exact same reasoning is what has resulted in the massive archive bloat
that Debian (and now the MOTUs) are forced to deal with.  Once a package
goes in, it's far more difficult to get rid of it again.  You've got to try
and make sure that users know that the package is no longer available
(aptitude/synaptic doesn't make them particularly aware, AFAIK), and there
will almost certainly be stink from someone when a package goes away.

If you try to avoid that problem by never removing a package, you're doing
users a disservice because they can have *no* idea whether a package they're
installing is up-to-date, has crippling bugs, or is going to leave their
machine horribly insecure.

Many packages with similar functionality is also a problem from the
point-of-view of splitting users and developers -- instead of making one
tool really good, with lots of user feedback and advice, you have 25
equivalent tools, none of which are particularly good because none have all
of the bug reports and developer interest.  Yes, FOSS is about choice --
it's also about allowing you to make the choice and say "no, we're going to
encourage people to use the better tools and not encourage proliferation of
*unnecessary* diversity.

Saying "we'll remove it if it becomes a problem" means you have to have some
way of tracking what to remove, when, and silencing the screams from the two
people in the world who feel like complaining about the problem, but without
actually volunteering to help solve it.  Let me tell you, from experience,
that identifying packages to remove is a very hard problem.

On the other hand, if you're happy for Universe to really just be a dumping
ground for all possible packages, with some sort of auto-removal capability
if they fail to meet certain criteria (better you than me to work out what
the criteria are, though) then you *could* just let any upload through --
but the fact that you're trying so hard to keep the quality level up through
REVU, mentoring, and pro-active work on packages, suggests that you're *not*
happy for "Ubuntu Universe" to mean "every possible package, take pot-luck
on whether it sucks or not".  But part of that quality evaluation is "does
this new package improve the distribution?", and I doubt that the 15th
script to run rsync for a backup is really improving the distribution.

- Matt