Even though the idea of formal alignment between the freezes of Debian and Ubuntu didn’t hold, there has been some good practical collaboration between the maintainers of key subsystems. There are real benefits to this, because maintainers have a much more fruitful basis for sharing patches when they are looking at the same underlying version.
Harmonization for Ubuntu 10.04 LTS and Debian Squeeze
I think this is where we stand now:
|Kernel||2.6.32 + drm-33||2.6.32 + drm-33||2.6.32||2.6.32|
|Mono (thanks Jo Shields)||2.4-snapshot||2.4-snapshot|
I’m sure there are inaccuracies, please help me keep this up to date, sabdfl on freenode is the best way to reach me. The RHEL and SLES numbers are third-hand, so up-to-date information would be appreciated.
The actual release dates of Ubuntu LTS and Debian will vary of course, because of different priorities. And there’s no requirement that the same base version be used for every major component – there may well be differences allowing for different approaches. But where we do have it, we’ll be able to collaborate much more effectively on bug fixes for key upstream pieces. If a lot of distributions pick the same base upstream version, it greatly increases the value of extended shared maintenance and point releases of that upstream.
Why every two years?
Two years is a compromise between those who want 1 year releases for better support of cutting-edge hardware and those who want 7 year releases so their software stack doesn’t change before their job description does ;-).
A whole-year multiple has several advantages. It means we can schedule the processes that are needed for collaboration at the same time of year whenever we need them – unlike 1.5 or 2.5 year cycles. Three years was felt to be too long for hardware support. Two years is perceived to be the Goldilocks Cadence – just right.
What are the criteria for choosing a common base version?
In both the Ubuntu and Debian cases, we’ll be making a release that we support for many years. So be looked for versions of key upstreams that will pass the test of time. Sometimes, that means they can’t be too old, because they’ll be completely obsolete or unmaintainable in the life of the release. And sometimes that means they can’t be too young. In general, it would be better to be reviewing code that is already out there. But there are also lots of upstreams that do a credible job of release management, so we could commit to shipping a version that is not yet released, based on the reputation of the community it’s coming from.
What if there’s no agreement on a particular kernel, or X or component-foo?
We will almost certainly diverge on some components, and that’s quite OK. This is about finding opportunities to do a better job for upstreams and for users, not about forcing any distro to make a particular choice. If anyone feels its more important to them to use a particular version than another, they’ll do that.
It’s really helpful to have upstreams and other distributions participate in this process.
If you’re an upstream, kick off a thread in your mailing list or forums about this. Upstreams don’t need to do anything different if they don’t want to, we’ll still just make the best choices we can. But embracing a two year cadence is the best way you have to be sure which versions of your software are going to be in millions of hands in the future – it’s a great opportunity to influence how your users will experience your work.
Of course, we’d also like to have more distributions at the table. There’s no binding commitment needed – collaboration is opportunistic. But without participating in the conversation one can’t spot those opportunities! If you represent a distribution and are interested, then please feel free to contact me, or Matt Zimmerman, or anyone on the Debian release management team about it.
I think this is a big win for the free software community. Many upstreams have said “we’d really like to help deliver a great stable release, but which distro should we arrange that around?” Upstreams should not have to play favourites with distributions, and it should be no more work to support 10 distributions as to support one. If we can grow the number of distributions that embrace this cadence, the question becomes moot – upstreams can plan around that cycle knowing that many distributions will deliver their work straight to users.