Re: Debian decides to adopt time-based release freezes
On Thu, Jul 30, 2009 at 11:49:48AM +0200, Steve Langasek wrote:
> Doesn't this imply that everyone who continues using Debian today does so
> merely as an accident of the release schedule and the particular set of
> packages that land in a given Debian release?
That and the fact that upgrades between Debian stable releases are easier
(or, at least, more officially supported) than from Debian to Ubuntu.
At the moment I could recommend Debian stable over Ubuntu LTS because
it has more recent packages (2009/02 release versus 2008/04 release),
or because it's an easier upgrade for people with existing Debian systems.
With synchronised releases, both those reasons to run stable disappear.
> OTOH, perhaps you're saying that you think that the proposed sychronization
> will be successful, and as a result Ubuntu's quality will come up,
> eliminating a key differentiator between the two at present?
I'm not aware of any apples-to-apples comparisons of Debian's and Ubuntu's
"quality"; but personally I haven't seen much evidence that Debian's
is significantly superior (NB: I haven't used Ubuntu LTS personally,
though). The tradeoffs to me seem to be:
Debian stable Ubuntu LTS
2 year rel cycle 2 year rel cycle
3 years security 3 years desktop security, 5 years server
guaranteed freeze date guaranteed release date
support for all pkgs support for main, best-effort for universe
stabilise from testing upgrade support from previous Ubuntu 6mo release
upgrade from oldstable upgrade support from previous Ubuntu LTS (?)
support for 6-12 archs support for 2-3 architectures
availability of pre-installed systems
full-time security support staff
commercial quality support
some additional packages
Having stable and LTS have mostly the same packages makes apples-to-apples
quality comparisons easier, which might be good or bad for Debian
depending on what the difference is. It'll make cross-grades from Debian
to Ubuntu fairly easy, removing most of the lock-in on Debian's behalf; and
For otherwise unsupported packages in Ubuntu universe, any security
problem that Debian notices can be copied straight into Ubuntu due to
synced package versions, making "best-effort" mean "at least as good as
Debian", so there's no drawback to using packages in universe.
So afaics, Ubuntu LTS looks to be the better system to use in all but
niche cases (non-x86/amd64 machines).
> There seems to be an assumption here that Ubuntu would benefit from bugfixes
> from Debian developers, but that the reverse would not be true. Is this
> what you believe? Does that mean you don't think Ubuntu developers
> contribute fixes back to Debian today?
Ubuntu has a well-defined and efficient process for accepting changes
from Debian (pull from unstable regularly), Debian doesn't have a
similarly efficient process for getting contributions from Ubuntu
(Ubuntu folks file a bug, maintainer eventually incorporates it), and
that'll presumably be made worse if there's a Debian freeze for most of
the LTS development cycle. So yeah, I think it's reasonable to expect
Debian won't get that many benefits from work on Ubuntu LTS into the
corresponding stable release.
Testable/refutable claim: my impression is most changes developed for
an Ubuntu release don't make it into Debian testing/stable until after
that release is out.
I'm not particularly bothered by this in and of itself -- if Ubuntu
LTS becomes better in every way than Debian stable is now, well great:
let's all use that instead! Benefits of free software, etc! But if stable
doesn't get used much because LTS releases (or short-term-support Ubuntu
releases) are way better, I expect that will have a flow-on effect
making testing and unstable less useful/effective, which in turn will
make Ubuntu less useful/effective. That doesn't sound like a fun outcome
for anyone to me.