March 12, 2007

Candidates for Debian Project Leader sound off

Author: Joe 'Zonker' Brockmeier

Once again, the Debian project is gearing up to elect a new project leader, with voting set to begin late this month. As we did last year, we asked the DPL candidates to sound off on some of the issues that will face the Debian Project in the next year.

Out of nine candidates, six took the time to respond to our questions via email. Steve McIntyre, Sven Luther, and incumbent DPL Anthony Towns failed to respond in time for this article. We received responses from Wouter Verhelst, Aigars Mahinovs, Gustavo Franco, Sam Hocevar, Simon Richter, and Raphaƫl Hertzog.

Though they didn't talk to us, Towns and McIntyre have provided platforms that are published on the Debian vote Web site. Luther has been relatively quiet on the debian-vote list, and has published no platform.

What's important for Debian?

We began by asking candidates what the most pressing issue for Debian is over the next year. We received an interesting range of responses. Franco says that, after Etch is released (presumably sometime in 2007), Debian should "make sure we get the Lenny release planning right from the start and work to switch for a serious goal-based release schedule." To that end, Franco suggests "we should define what are the key components that need to be ready, and push the other components (more than 15,000 packaged softwares) to follow this approach. Those who fail won't be released."

As in years past, several prospective DPLs say that one of Debian's major issues is communication. Hertzog says that communication is the top priority for the next year. "Too many developers have lost confidence in some key teams because they communicate badly and they don't seem to accept help nor seek for it.... Most of the time the consequences are not very big, but it's terribly annoying to have to run after other people to be able to go forward with your own projects."

To solve the communication problem, Hertzog says that he would "work with the problematic teams to enhance their process and to actually recruit and train new volunteers."

Richter's response displays concern about the health of Debian itself, saying that the most pressing issue is "avoiding the project falling apart." Again, Richter's concerns touch on the issue of communication. "The project has reached a size where it is impossible for everyone to know everyone. I'm doing my best, but I have close to zero contact to the South American people, for example. As a result subgroups are forming, based of geography or common interest. These usually get along well, unless you add in some thing they feel they have to compete for, like time slots at a conference or, in the Dunc-Tank case, money."

Hocevar also cites "insufficient internal communication" and discouraged developers. Hocevar says he would appoint additional people to core teams in Debian, which would "get things done faster and by requesting proper reporting, which will force people in place to either explain their refusal to do things or let others do the job."

In addition, Hocevar also says that Debian has "lost user appeal." According to his platform, Debian needs to be "sexy again."

"Even the FreeBSD Web site is sexier than ours. How can we expect to attract users? Have you seen how the Web site renders on Internet Explorer? We lost countless users to Ubuntu. Admittedly, Ubuntu drew a lot of its users from other distributions and even from Windows. But there is no reason for Debian to be simply 'the distribution upon which Debian-based distributions are based.' We can also get new users by being more appealing."

For Mahinovs, the objective for the next year would be "to cool down and think forward." He says that Etch should be released within the year, but that "faster and faster release cycles" have brought stress to Debian development. "I would suggest only releasing Etch within my term and then discuss long-term projects to move Debian into unexplored territories, like the couple of ideas in my platform."

New blood

A frequent complaint about Debian is the length of time it takes for prospective developers to become full-fledged Debian Developers (DD) with voting rights and upload privileges. Waits of more than a year are not unheard of. We asked the candidates if this was something they'd try to address as the DPL.

A few candidates took the position that a lengthy wait was desirable. According to Richter, "the length of the process is actually a good thing. Being an 'official' developer has two implications: the right to vote, and the right to sign uploads for processing by the archive maintenance software. Not being a DD does not stop you from doing uploads under your own name, it just means someone else will have to look over your stuff first.

"This means you have to actually go out and convince someone to look at your package and upload it afterwards. Normally, you will end up with a few 'default' sponsors, whom you keep contact to and ask questions, and you have at least one common topic when you meet in real life. Thus, the process helps form the community by establishing mentoring relationships; the fact that it is a long process turns it into some sort of initiation ritual that helps forming a bond with those parts of the community you haven't met. This notion is strengthened even more by receiving a formal right (voting) that you didn't have before and could not even execute by proxy."

Mahinovs says that one year is a reasonable amount of time to go through Debian's new maintainer process. "Debian Developers in effect have root access to any Debian machine out there, so we need to be very careful when evaluating potential candidates, including evaluating their patience, which will be useful in the mailing list discussions.... However, as there are case when the NM process takes more then one year, it should be looked into from time to time."

Franco, on the other hand, says that the process should not take more than a few months, and suggests that the team handling new maintainers should be granted additional resources to help speed up the process. Hertzog says that the length of time to become a full-fledged developer is not a problem, but that it should be easier for non-DDs to be granted limited upload rights.

Dunc-Tank and funding

The Dunc-Tank project, which sought to pay Debian release managers to push Debian Etch out the door on schedule, stirred up a fair amount of controversy among Debian Developers and led to an unsuccessful attempt to remove Towns from the DPL position. Ultimately, Dunc-Tank failed in its objective, and Etch remains in development.

Hertzog, who was on the Dunc-Tank board, says that the experiment taught developers "we can get enough money to run some projects, but it didn't give us a way to do it without causing too much trouble." However, Hertzog still supports the general idea of an "open funding structure" for Debian, and says "it's still a long-term project for me to find out how such a structure could work, but it wouldn't be a priority of mine if I'm elected. We have more urgent stuff to do."

Rather than raising money via a mechanism like Dunc-Tank, Hocevar says he would endorse projects like Google's Summer of Code. "My main concerns were about where the money came from, who decided where it went, and how the priorities were decided. There are ways of funding Debian developers that have never met any violent opposition, for instance the Google Summer of Code." He also says that there's a "gray area" between funding mechanisms like the Summer of Code and Dunc-Tank.

Verhelst says that the has "no issue with paying Debian Developers per se," but doesn't think "creating a system to decide by committee who gets money and who doesn't is a very smart thing to do."

Many candidates suggested that it would be better to fund development within Debian by external means. Richter says that the Dunc-Tank method was "the wrong way to go," but says that external groups sponsoring work on Debian should be fine. "For example, I can see a computer manufacturer sponsoring a port without anyone thinking that their group would have been sponsored if they had done things another way."

According to Mahinovs, funding "must not flow trough Debian or SPI in any way and there shall be no guarantees from the Debian project that the work of such paid individuals will be incorporated into Debian. Money brings pressure, and I think that in large-scale contributions volunteers should have a way to review changes made by company-hired developers to guard against any potential abuse."

Franco says that he would not attempt another project like Dunc-Tank, but would "seek funds for the Debian project as a whole and concentrate more on event sponsors."

Election timeline and procedures

For its DPL election, Debian uses the Condorcet method, where voters rank their choices rather than voting for a single candidate. Details on Debian's voting method are available in its constitution.

The campaigning period for the DPL post runs through March 18, and voting runs through April 8. According to the Debian Vote page, 1,013 Debian Developers are eligible to cast ballots in this election.

Click Here!