Kubuntu and main/universe

[Scott Kitterman]

On Wednesday, May 23, 2012 12:49:54 PM Colin Watson wrote:
> Hi Kubuntu folks,
> 
> At this UDS, the old "archive reorg" topic came up again
> (https://wiki.ubuntu.com/ArchiveReorganisation for those unfamiliar with
> it; this page also contains some background for some of the Ubuntu
> archive jargon I've inevitably ended up using in this post), and I had a
> side discussion with Jonathan about a Kubuntu-related question arising
> from it.
> 
> Ultimately, I still do plan to move all free software in the archive to
> a single consolidated "main" component and express who supports what in
> ways that are rather more flexible than (effectively) moving files
> around in the pool; but it's clear at this point that I'm only going to
> be able to lay the groundwork for that this cycle, not actually do it.
> In advance of that, though, my question is as follows: is being in main
> actually much more than an obstruction to Kubuntu development these
> days?
> 
> Up to now, the packages that make up Kubuntu - at least the images you
> ship - have been in main at least in part because Canonical was offering
> some level of support for them, and thus we wanted to make sure they
> passed an explicit review by our security team and the like.  Now that
> that isn't the case, though, being in main basically means that
> Canonical is requiring main inclusion review for Kubuntu even though we
> aren't offering support for it, out of step with all other community
> flavours, and thus imposing an extra chunk of work on you that nobody
> else has to do.  That seems rather unfair!  When I spoke to him at UDS,
> Jonathan seemed to feel the same way, and indicated that you might well
> want to consider moving Kubuntu-specific packages to universe.
> 
> Now, this is one particular way to look at things, and I realise that
> you could cast the question differently.  Since the main component has
> that extra level of review associated with it and has a more restricted
> uploader group, many people do regard it as having an extra claim of
> quality.  I don't want to be the guy "kicking Kubuntu out of main"; if
> we make this change in advance of a more sweeping archive
> reorganisation, I'd much rather do it with your consent and indeed
> enthusiasm.  My conversation with Jonathan suggested that this might
> well be forthcoming, but I said I'd mail kubuntu-devel to see what other
> people think.
> 
> Some other details:
> 
>  * What effect would such a change have on upgrades?  update-manager
>    shows a "no longer supported by Canonical" message for installed
>    packages that have been moved from main to universe between releases;
>    does muon have anything like this?  I don't want to scare off Kubuntu
>    users.

Which, depending on what 'supported' means is accurate (see below).

>  * Moving to universe would imply not using language packs any more,
>    because uploads to universe don't have translations stripped out and
>    sent through the Launchpad translations system.  Jonathan suggested
>    that this would be a desirable change for Kubuntu, but I expect that
>    at the very least it would be a respectable chunk of work for
>    somebody to unwind all of that infrastructure.  (As an aside, we'll
>    need to think about how this would work post-archive-reorg if more or
>    less everything ends up in main.)

This is the biggest plus associated with moving to Universe, IMO.  The 
infrastructure is fragile enough that I don't think it's particularly more 
work to unwind it than keep it.  Since upstream KDE already ships per language 
translations the primary advantage with the LP language system isn't 
applicable to KDE.

>  * A number of Kubuntu-specific packages would need to stay in main for
>    transitive-closure reasons: source packages in main that, for
>    example, build bindings for both GTKish and Qtish worlds will
>    generally build-depend on both, which will cause those
>    build-dependencies to stay in main along with all their
>    (build-)dependencies.  I don't want to encourage people to start
>    splitting source packages for the sake of this, so a number of base
>    packages would stay in main regardless.  Still, lots of UI packages
>    would be able to drop out to a component with less restrictive
>    processes.
> 
>    I've attached a file showing a first pass at what the results would
>    have been if we'd done this for 12.04.
> 
>  * Are there any effects on mirroring that you care about?  Right now,
>    it's in principle possible to run an archive mirror useful to Kubuntu
>    users with only main (and maybe restricted).  This change would mean
>    you'd have to have universe as well.  On the other hand, Xubuntu et
>    al have this problem today anyway, and maybe you only care about
>    things like the major per-country mirror network which carries
>    everything.
> 
>  * Would there be any "soft" questions of PR and the like that would
>    need to be addressed?

I suspect we've hit those already, so it's not a major driver.

The benefits to staying in Main, as I see it:

 - Canonical security team support (if they aren't willing to support, then I 
think we're pretty much by definition out of Main)

 - Some PR advantage

 - Limited set of non-Kubuntuish uploaders.  There are some people with MOTU 
upload rights that I wouldn't want going anywhere near Kubuntu packages and so 
I see it as a positive feature that they can't.  The kubuntu-dev process is 
responsive enough that it's easier for someone working on Kubuntu to get that 
then MOTU, so in terms of Kubuntu related upload rights, there's no particular 
advantage moving to Universe.

 - Build priorities: Thanks to Colin's recent efforts our seeded packages would 
still build after unseeded Universe packages, but would build after everything 
in Main/Restrictred is done.  At busy times (and when we're trying to get a 
new KDE SC release uploaded, built, and announced) this is going to have a 
significant impact on how long it takes to get stuff done.

Other than security support (and I would assume the availability of commercial 
support contracts, but I don't actually know about that) there isn't a lot of 
post-release support for Kubuntu specific to it being a supported flavors.  
We're doing post-release microversion KDE SC updates, for example, and those 
are all done by the Kubuntu community.

There are some things we could do in Universe that we can't in Main and we 
avoid the work of the MIR process, but if things are not going to get through 
the MIR process, it's probably just as well we don't ship them by default.  I 
think the process has enough value that it's not just work avoided if we don't 
have to do it, we lose a significant QA point as well.

I have assumed we're going to have to move to Universe, but have a preference 
not to if we have a choice in the matter.

Scott K