Debian leaders: Faster release cycle required

Debian leaders: Faster release cycle required

Summary: The lack of a new stable release of Debian GNU/Linux since July 2002 is fuelling the campaigns of many candidates for the project's leadership position, with many pushing for a shorter and more regular release cycle to arrest user departures. According to one developer and Debian Project Leader (DPL) candidate, Matthew Garrett, Debian is not releasing a stable version of its software often enough.

TOPICS: Open Source, Linux
The lack of a new stable release of Debian GNU/Linux since July 2002 is fuelling the campaigns of many candidates for the project's leadership position, with many pushing for a shorter and more regular release cycle to arrest user departures.

According to one developer and Debian Project Leader (DPL) candidate, Matthew Garrett, Debian is not releasing a stable version of its software often enough. In a recent statement outlining his election platform, Garrett said: "At the time of writing, it has been 31 months since the last Debian release in July 2002. This is not acceptable. The release team has been hampered by communication issues between teams, resulting in release-blocking issues not becoming apparent until they are too late to avoid."

The DPL is the organisation's official representative. Duties include representing the project to external bodies as well as managing the project and defining its vision. The position is an elected one (all Debian developers are eligible to vote and run for office) and has a one-year term. This year's voting period for the elections will close on 11 April. The current Debian project leader, Martin Michlmayr, is not standing for the post again.

Garrett recognised the severity of the situation in his statement, saying that if he was elected he would "liaise with the security team and stable release manager to ensure that our users are not left needing to upgrade excessively or left running a distribution that no longer has official security support."

Garrett is not alone in expressing his sentiments. Although they are all competing for the votes of the currently 945 Debian developers spread around the globe, Garrett's fellow candidates for the DPL position seem to agree with him. Jonathan Walther, who has been involved with Debian since 1998, has based his entire running platform around his opinion that Debian should "release early, release often".

In his statement, Walther said that Debian's slow release cycle has resulted in the fact that "many users have been leaving Debian, or choosing other distributions without even trying Debian first". If he was elected DPL, Walther said, "my first focus will be on our release schedule. I have a proven history of releasing software on time, on schedule."

Developer and candidate Andreas Schuldei, who is paid by his employer to work on Debian, said in his own statement that the current delays in the release schedule were causing frustration and a decline in morale in the Debian community. Schuldei, too, would focus on a speedier release cycle if he was elected.

Red Hat general manager for Australia and New Zealand Gus Robertson recently told ZDNet Australia  that Red Hat was locked in to a 12-18 month release cycle with Intel, IBM and Oracle, in stark contrast to Debian. Debian's last two official releases, code-named 'Potato' and 'Woody', were released in August 2000 and July 2002 respectively.

While Debian does make more updated software available through the Testing and Unstable versions of its distribution, the Debian Project makes it clear on its Web site that the versions are not suitable for production use. The Testing distribution is described as "not completely tested ... [with] no official support from the Debian security team", while Unstable should be run by "developers and those who like to live on the edge".

Comments from the other three DPL candidates were generally not as inflammatory with respect to Debian's release cycle, however at least two candidates, Angus Lees and Branden Robinson, have expressed their support for a recent proposed list of changes to the release cycle that came out of a face-to-face meeting of those involved in release management. The proposal, which Robinson is calling the "Vancouver Prospectus", proposes making some radical changes to the Debian's Linux offering.

Perhaps the most drastic proposed change is for Debian to stop supporting some hardware architectures (11 are currently supported). Announcing the proposal, developer Steve Langasek said: "The reality is that keeping eleven architectures in a releasable state has been a major source of work", and that it is hoped that the cuts will give Debian "a much shorter release cycle on the order of 12-18 months".

If the proposal ends up being approved by the Debian community, Debian will end up supporting only the i386, PowerPC, IA64 and AMD64 architectures and dropping several architectures such as Sun Microsystem's SPARC and IBM's S/390. The changes would not affect the next Debian release (code-named 'Sarge') but would be implemented for the following release, code-named 'Etch').

The proposal reflects comments from Linus Torvalds last week that he personally believes the x86-based 64-bit and PowerPC platforms are the two most relevant architectures being currently used.

Another possible solution for the organisation is to set in stone a regular release cycle based on a set time period. Walther, at least, is openly in favour of a 6-month cycle similar to that used by the OpenBSD project, which he told developers in an online debate this week that "has developed the optimal release strategy". Fellow DPL candidate Anthony Towns agreed in the same debate that Debian's release cycle should be "predictable to the day", although he proposed varying intervals between releases from 6-18 months.

No final conclusion has been reached yet by Debian on how to shorten the release cycle, but it is possible that the new leader will drive a greater clarity of vision after the results of the DPL election are announced.

Topics: Open Source, Linux

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Debian's major issue is that they want to do it all (and with no money).

    I just had a look at the "testing" Packages file on one of my machines: it currently lists 14747 Packages. Does that make any sense? By the way, downloading and parsing this huuuuge file each time you want to perform a minor update already rules out a lot of machines, so the drop of some "architectures" is not so new in a sense.

    Modularizing debian would enable debian developers to focus a "core" debian (yes, just like the realistic Fedora "core" distribution) and to release it more often. The submodules would follow if they can.

    Ubuntu which is based on debian sort of got it since they splitted this huge list into a "main" component fully supported, and the rest.

    This 14747 package list is especially disappointing considering that Debian's package management tools are already able to handle multiple sources in a seamless way.
  • Is Debian now to big to focus ? From outside, it looks like debian is a bit of a bureaucracy with many organisational layers or worse, silos.

    About release, Debian should split between server and desktop releases. Secondly they shoud do more Snapshots from unstable. Too many people use unstable (for desktop) to ignore.

    But maybe that's the problem with Debian : It doesn't matter what anyone says, Debian has its ways and that's it.
  • Other architectures are NOT being dropped! What is being discussed is releasing "stable" releases only for the "first-cl****" architectures. The others are still being supported and worked on, not just required to sync with the stable releases. One of the big problems with the stable releases has been managing all 11 architectures to sync, when in fact it's doubtful whether ANYBODY uses the stable release for some of these architectures.
  • As a user of debian stable on servers, I have no desire to see them roll out new stable releases every 18 months, because I have no desire to constantly upgrade to newer versions on servers that don't require it. What I suggest, is to have stable server and stable-desktop, with stable server they should make a part of setup so those that needed newer packages for something like woody can get them, but upgrading the whole distro just to attract new users is the wrong reason. Thats why they could have debian-desktop and roll over regularly, leave us with our stable servers.
  • I chose Debian this year for a production system because it has so many packages and because Debian is so fussy about releases. I want quality software.

    Linux has problems with installing from source because of incompatible libraries. As long as the number of libraries out there stays huge, I need a huge distro like Debian to be more sure of installing applications.

    Eliminating the requirement to release all ports simultaneously may speed things up for my favourite ports but it may allow bugs in some ports to go unfixed. The only urgency I feel is an AMD64 release. I can live with i386 for the time being with only a small penalty in performance. Next year, I may need increased performance, but I can always add a few extra servers. I use Linux Terminal Server + Debian so I need all the features of Linux as a great server paltform, the flexibility of the Linux desktop, and the reliability of Debian. Getting all this from Debian is worth the waiting...

    Robert Pogson
    beautiful, downtown, Deline, on Great Bear Lake, Canada
    home of hockey and the best fishing in the world.