The 2004 Debian Project Leader election [LWN.net]

March 24, 2004

This article was contributed by Joe 'Zonker' Brockmeier.

It's election time again. The Debian Project is holding its annual election for Debian Project Leader (DPL). This year, three candidates are running for the office: current DPL Martin Michlmayr, Gergely Nagy, and Branden Robinson. Debian Developers also have the option of voting for "none of the above" if they prefer.

We contacted each of the DPL candidates with several questions about themselves and their intentions in running for office. We also combed through the discussion on the debian-vote list, where the candidates have been participating in discussions about the Debian project, and why they are qualified to be DPL -- or why they are not. We have attempted to distill all of this information into a brief summary of the candidates' platforms and ideas, but we recommend that LWN readers interested in the DPL election also take the time to read each candidate's platform (they are all available on this page) as well as the relevant DPL threads on debian-vote.

It's typical for candidates for any office to assure their voters that they take that office seriously. Not so with Nagy, who it seems is running on a whim. Nagy is a 22-year-old student living in Hungary, who is running "for fun and profit, of course!"

Past DPL elections were too serious for my taste, too much political stuff, and not much fun. For me, Debian is a hobby, and a hobby should make me laugh at times. That is what I intend to do by running and giving nonsensical answers to otherwise good questions.

He asks Debian Developers not to "even think about voting for me," and says he would resign immediately in the event he does win. Michlmayr and Robinson are a bit more serious about the election.

In addition to serving as DPL, Michlmayr notes that he is also working on a Ph.D. at the University of Cambridge. He says that he is researching quality management in free software. Michlmayr already holds Master degrees in Philosophy and Psychology from the University of Innsbruck, and a Masters in Software Engineering from the University of Melbourne. Michlmayr told LWN that he is running for a second term as DPL to continue his work:

Due to the size of Debian, the project requires a lot of coordination and leadership in order to keep the project running smoothly. While we have a high number of excellent developers and package maintainers, few people are interested in or have the skills to coordinate the project. I have been involved in coordination activities for many years, and think that this is the area where I can contribute most. I have acted as Debian Project Leader for almost a year now, and feel that I have done a good job. I would like to continue my work, and thereby make sure that the project runs smoothly and that other people in the project can carry out their work. I'd also like to continue representing the project to the outside, by attending conferences and talking to companies.

The kind of tasks I carry out as DPL are summarized in my "6-month retrospective."

Robinson lives in Indianapolis, Indiana and has worked for Progeny for the past three and a half years. He has been a Debian Developer since 1998, and served as Treasurer of Software in the Public Interest (SPI) from August 2001 to February 2004. Robinson points out in his platform several reasons why he is running for DPL. Robinson writes that Debian needs improved, more open, and more visible processes. Robinson also says in his platform that the Debian Project should "take our Constitution more seriously," and that the Debian project needs "a leader who will champion our cause:"

Debian is making inroads, seemingly everywhere; I want to accelerate that process and evangelize Debian everywhere I can. I don't see the phenomenon of subprojects or compatible forks as a threat to us at all; instead, it is a beacon of our success. It's my opinion that it is within our power to make Debian a de facto industry standard; the company I work for achieved certified LSB compliance for a snapshot of Debian "sarge" in January. I was enthusiastic about Debian from the day I became a maintainer, and I'm still excited today. Furthermore, I can effectively communicate that enthusiasm and excitement to an audience.

Since the DPL serves a one-year term, we asked each of the candidates to identify the biggest challenge facing Debian over the next year. We also asked candidates to rate the "health" of Debian, and whether the "market share" of Debian as a Linux distribution was a concern. Michlmayr responded:

I think market share is important, and the recent Netcraft survey showed us that Debian is doing very well. One of the big challenges will be to adopt a faster release cycle, and to support current hardware better. We also increasingly have to work with companies, to get better support for Debian (commercial support, hardware support, having Debian pre-installed).

In his platform, Michlmayr also lists several goals he has for the next year. In addition to a faster release cycle, he says that Debian needs a clear release plan for the coming release and for the release cycle for the next few years. He also cites a desire to work with external projects to help reduce duplication of effort between Debian-based distributions.

Robinson told LWN that he sees scalability as the top problem for Debian in the next year:

The biggest challenge facing us is our answer to the question "how can we scale?" We're huge -- over nine hundred developers, at least half of whom are active enough to have participated in the "non-free" General Resolution process, which means we probably have on the order of four to five hundred reasonably active developers. Even that figure dwarfs the engineering staff of all but the largest software companies.

We're also huge in terms of distribution. The Debian "sarge" release is anticipated to consume 13 CD-ROMs' worth of space for the x86/IA-32 binary packages alone... We're also big in terms of infrastructure. We have, at present, 35 project machines in our LDAP database. This doesn't list many quasi-official machines, such as many in the build-daemon network which keep our packages built for all eleven of our architectures. Just about any serious Linux user can imagine how much work it would be to keep that much hardware up and running; an experienced sysadmin knows of whole new dimensions to the problem. Add to that the fact that in many cases, our top-tier administrators don't have easy physical access to these machines, and the scope of difficulty is magnified again.

As for market share, Robinson said that it is something that the DPL "should be cognizant of, though his or her ability to directly affect it is almost nonexistant." Robinson also said that he is "not too worried," about the relative market share of Debian and that he "cannot help [but] be aware of the rising tide that is Debian."

Debian Developers recently rejected a proposal to remove non-free. However, each of the candidates for DPL says that they support removal of non-free from Debian. Robinson said that non-free software does not directly serve the Debian mission, and pointed out that many voters may have misconceptions about the nature of non-free:

Time and again during the long discussions leading up to the vote, the preservation of the non-free section was defended on the grounds that it would take packages away from users -- often using as examples packages which weren't actually *in* the non-free section, hadn't been for years, and for which there was no reasonable expectation of return.

Advocates of dropping non-free, like myself, need to do a better job of dispelling this sort of fear and ignorance, so that people who favor its retention at least can do so on informed and rational grounds. If we do, at some point in the future when the issue is revisited, if the proposal fails again, it will at least do so based more on its actual shortcomings, rather than imaginary horrors.

Michlmayr also wants to get rid of non-free, and points out that as long as Debian maintains non-free that it is less likely that free software alternatives will be created to replace the non-free packages. He said that he was not surprised by the vote, because "the non-free removal was not approached properly."

The non-free opponents simply wanted to remove the non-free packages, but did not offer a transition plan. While there has been talk of moving the non-free packages to an APT repository on non-free.org, nobody has done so yet. In the interest of our users, I think we should first move non-free packages to an outside project, help them get started, and mirror their packages on our mirrors for a year or two to let users switch to the new APT repository...At that point, we can stop distributing those packages ourselves.

Another issue that comes up from time to time is Debian's support for multiple architectures. We asked the candidates whether support for multiple hardware platforms was slowing the project, and when Debian should consider dropping a hardware platform. Nagy responded that if any architecture were dropped, "it should be x86, period...Debian being the Universal OS, should support all possible architectures, and as long as there are people who do the porting work, the support for the platform must be kept."

According to Robinson, the answer should be to improve the build infrastructure:

If an arch proves to be unsustainable, I think we should probably officially discontinue it rather than move it into some sort of "slow lane". If there aren't enough people dedicated enough to keep the port alive in Debian, I suspect there won't be enough people to keep it alive *outside* Debian, either.

In his response to LWN, Robinson also said that Debian should stop supporting a platform "when our developers are no longer able to maintain it to our standards." According to Robinson:

That some architectures take days to compile packages that on modern CPUs take only hours is, interestingly enough, less of a real problem than packages that slip through the cracks and don't get built at all.

Michlmayr also said that support for multiple hardware platforms was not the cause for slow releases:

Supporting the number of platforms we do is certainly a challenge, but it is actually not the main reason we're sometimes slow. I think the community benefits from our wide support of platforms, since we report lots of toolchain bugs (GCC, binutils) on many architectures; we also support some architectures nobody else supports, and it would be a pity if nobody supported the[m] anymore. One reason why Debian has slow releases is the number of packages, and that some of these are not well maintained. This is an issue we have to approach, possibly by moving to maintainer teams rather than relying on a single maintainer for a package.

Finally, we asked the candidates about their thoughts on projects that make use of Debian, such as Progeny's "Componentized Linux," and Bruce Perens' UserLinux, and whether companies like Lindows.com and Xandros were giving enough back to the Debian Project.

Michlmayr said that he has contacted some of the companies that make use of Debian, and that he thinks that "closer cooperation is very important." He notes in his platform that there is limited cooperation between the Debian-based distributions, and that there is development that is not being integrated back into Debian.

As the Debian Project Leader, I would see it as my duty not only to work with these external projects, but to try to internalize them as much as possible. This is partly happening already, but I'd like to work with other projects more closely to drive this process along. As an example, Skolelinux (who have always contributed their work to Debian) first adopted our debian-edu project and are now moving towards using the debian-edu name as their brand. Furthermore, after discussions with developers of DeMuDi (a multimedia distribution based on Debian), they agreed to join our debian-multimedia project and to merge their work into Debian.

Debian will benefit to a great degree if more Debian based projects get involved and make contributions. I am very excited about this because many of those projects are sponsored by local governments. Just imagine the great advances we can make if there are a few paid people in countries like Brazil, Greece, Norway and Spain (which are all working on Debian based distributions). While I cannot control what those projects do, I intend to work together with them as closely as possible. Everyone will profit by more cooperation, and I am interested in helping with the coordination to make this possible.

Robinson responded that one problem presented by the many Debian-based projects, and the "vast amounts of Free and Open Source software that we see today," is that it's hard for people to determine whether the problem they're trying to solve has been solved already.

This isn't just a matter of finding out whether there's a Freshmeat, or SourceForge, or GNU Savannah project in the problem space. That's relatively easy. What's more difficult is finding out whether existing solutions are mature, robust, and a good fit for the remainder of your software (or organizational) infrastructure....

That, I think, is the challenge that Bruce Perens's UserLinux and Progeny's Componentized Linux initiatives are rising to meet. I don't believe it's any accident that two former Debian Project Leaders are among the first to appreciate this need. They witnessed first-hand the incredible breadth of the software prepared by the Debian Project, a breadth that has increased supra-linearly over time.

As to the question of whether companies give enough back to Debian, Robinson says, "yes and no."

I think these companies -- Progeny included -- do a good job of promoting the Debian name, and authoring freely-licensed enhancements to it. The challenge appears to be in coherent integration back into the Debian distribution itself...

I think this takes some initiative from both sides. In my Platform, I proposed officially delegating ambassadors or liaisons from the Debian Project to other organizations, and this can certainly include companies like Lindows and Xandros. At the same time, these companies need to be willing to pay someone to serve a complementary function on their end -- someone who will work with Debian and not let requests for information fall on the floor.

The current system, he notes, may be confusing for developers inside a company like Lindows or Xandros who wish to contribute but are unsure of the proper way to go about it. A Debian liaison to a company would serve as an interface between the Debian project and companies utilizing Debian and looking to contribute back to the project.

The DPL election will continue for a few more weeks. Debian Developers have until April 10 to cast their votes for the Debian Project Leader (DPL), give or take fifteen hours due to a snafu in sending out the call for votes. Good luck to all the candidates, and may the best developer win.

(Log in to post comments)