[Desktop12.04-Topic] GNOME Version for the LTS

[Michael Terry]

On Sat 15 Oct 2011 03:10:42 EDT, Jeremy Bicha wrote:
> Hi, I've got a strong opinion on this: I'm very skeptical about
> staying with GNOME 3.2. I don't think GNOME 3.2 is exceptionally
> stable or high quality, or that 3.4 will be exceptionally buggy. My
> blind guess is that 3.4 will generally have less bugs than 3.2 as 3.2
> was the first release to build upon the GNOME 3 transition. Only this
> fall will Debian, Ubuntu, and openSUSE users get GNOME 3 so several
> bugs haven't even been reported yet (and some bugs won't be fixed in a
> .1 or .2 update anyway). It would have been foolish to ship KDE 4.1 in
> two Kubuntu releases in a row for stability purposes. I believe this
> would be unprecedented for Ubuntu to skip packaging the latest and
> greatest stable GNOME (except for last spring which was a completely
> different situation).

I agree that 3.2 is not abnormally stable, nor that 3.4 will likely be 
abnormally unstable.  But each GNOME release does tend to focus as much 
on new features and rewrites as on bug fixes.  Features and code churns 
cause bugs too.  If they didn't, the number of GNOME bugs over time 
would go strictly down.

I used to work in the OEM Services team in Canonical.  We would take 
stable Ubuntu releases on a specific hardware platform, customize it, 
and fix bugs reported by the QA teams involved.  We also sometimes would 
take development Ubuntu releases and do the same thing, tracking Ubuntu 
development.

When working on a stable base, the kind of bugs you can work on are not 
"this feature doesn't work" but "in this corner case, the feature 
doesn't work", not "it leaks 1M a minute" but "it leaks 1K a minute", 
not "it crashes when I open it" but "it crashes when I press all my 
mouse buttons at once".  It's a matter of degree.

And of course, the Ubuntu desktop is more than just GNOME:  Unity, 
LightDM, Ubuntu One, Software Center, and hardware integration like 
multiple monitor support and bluetooth.  Those all would be able to get 
more stability attention too.

Holding back would make 12.04 less exciting and fresh.  But part of this 
question is "What does an LTS means to us?"  To me, LTS releases are 
what I should suggest to friends and family across the chasm.  People 
that don't want to upgrade every 6 months.  People that place a higher 
value on things "just working" than having the latest and greatest.

I hope that holding back could let us make 12.04 feel like 12.04.1, if 
you know what I mean.

> There's a vocal segment of the open source community who believe
> Canonical is forcing Unity on them and doing a terrible job at making
> GNOME available. Regardless of the (in)accuracy of that belief,
> deciding to stick with GNOME 3.2 will be a PR hit and we need to have
> a very easy-to-understand reason for that decision if it's necessary.
> I don't think GNOME developers would be very happy with the decision
> either and it's good to keep upstream as happy as possible.... :-)

I agree that holding back would create a messaging problem, in both a 
"not fair to GNOME Shell" and a "not an exciting release" sense.  But 
first, I think we have to decide if it's a good engineering decision.

As I said above, I'm happy if an LTS is not exciting.  And I think the 
GNOME concerns are misplaced.  Projects need people looking after the 
"long tail" of stability as well as new features.  The bugs we fix make 
it back to GNOME.  That was actually what I liked best during my time at 
OEM Services ( 
http://mterry.name/log/2010/09/15/what-i-do-in-oem-services/ ).

> The GNOME 3 PPA on Natty was honestly sort of horrible. Not to say
> that it didn't have benefits: I used it and for me and others it was
> quite nice to have. It was also good in encouraging new contributors
> to volunteer. And I am appreciative of the work it took to produce the
> PPA (which of course also really helped our GNOME 3 transition early
> in the Oneiric cycle). But it broke the normal Ubuntu desktop in quite
> a few unavoidable ways.

I hope that a 3.4 PPA would be less broken, because it's not such a 
crazy transition as 3.0 was.

-mt