Debian and time-based freezes [LWN.net]

By Jonathan Corbet
July 29, 2009

On July 29, a surprise announcement heralded a significant change in the way the Debian release process works. Rather than freezing the distribution when it was "ready," the release team will start to impose a scheduled freeze in December of every odd-numbered year, starting with December, 2009. There still will be no scheduled release dates, but the plan is to start the final phase of the development cycle in a scheduled manner.

It would appear that much of the Debian development community was as surprised as anybody else; there had been no discussion of this change on any of the project's mailing lists. The press release states:

The new freeze policy was proposed and agreed during the Debian Project's yearly conference, DebConf, which is currently taking place in Caceres, Spain. The idea was well received among the attending project members.

Many developers did not attend DebConf (which concludes on July 30), and those who were there disagree somewhat with the above description. It seems that some DebConf attendees, at least, feel that all they got was a few hours advance notice; the change was announced to them as something which had already been decided.

It should not be surprising that there is a fair amount of dissent in the ranks. This is Debian, after all. But there seem to be more than the usual number of complaints this time around. The key themes seem to be:

The change may or may not be good, but the way in which it was done was wrong. Debian developers should not learn about a major process change from a press release.
There is no reason to do a short development period to freeze this December when a freeze in 2010 would fit the two-year period perfectly. Shortening the "squeeze" development cycle halfway through will create havoc with many developers' plans and endanger a number of the objectives for the squeeze release. A lot of work will have to be crammed into the remaining time; some minor components, like the kernel, have not yet been updated.
Freezing in December will guarantee that Debian will ship obsolete versions of KDE, which releases in January.

The biggest grumble, though, appears to come from a feeling that the Debian project is being asked to change its ways and, arguably, compromise the quality of its releases for the sole purpose of accommodating the Ubuntu release schedule. One might dismiss this idea as overly conspiratorial, but it's worth reading this interview with Mark Shuttleworth, published on July 12:

And the really big news here is that we've been having very good discussions with the Debian release team. So the Debian release team has indicated that they are very open - not about a release date but a freeze date. That freeze date would be the time where we sit around and look at all the major components and decide what the major versions would be that we collaborate around.

In other words, Mark Shuttleworth knew about this change before the Debian developers - who are expected to implement it - knew. Given that Debian is supposed to be an open project, something which gives this kind of smoke-filled-room-decision feeling is guaranteed to be received poorly.

There are answers to some of these complaints, of course. Luk Claes, a member of the release team, said:

No, the Release Team proposed a plan. The project is free to accept or refuse the plan. Of course refusing the plan will have its consequences within the Release Team as well as within the project.

Even without the dark talk of "consequences," this statement will not have helped the situation; the press release says "Debian decides to adopt time-based release freezes," which is not the normal language used for a proposal. But it is true that the Debian release team is empowered by the project to make decisions like this. Meanwhile, the Debian press team claims that the abrupt announcement was required to keep journalists from mangling the news.

The short cycle is justified this way:

The main reason is that we now have the momentum to try a time based freeze and that delaying the freeze would cause developers to 'forget' about what a time based freeze is about.

The release team has also promised to talk with the Debian KDE maintainers to see what sort of solution can be worked out there.

But the release team has said nothing about Ubuntu and has not responded to the charges which have been made in that regard. It seems that a good case could be made for closer cooperation between Debian and Ubuntu - in fact, Debian developers have been asking for that for some time. Ubuntu has become a major (if not the major) distribution channel for Debian, increasing Debian's relevance in the process. If the combined distribution channel could be made to work better for everybody involved, the results should be good for both Ubuntu and Debian. It is hard to fault the release team for exploring ways to make Debian's release cycle work better for Ubuntu; one could, indeed, argue that it would be irresponsible for them to do anything else.

So the real question has to be: why has this conversation with Ubuntu been swept under the rug in the release team's communications with the Debian development community? It creates a strong impression of hidden agendas. The Debian project may now head into an extended period of more-than-usually acrimonious debate, dueling general resolutions, and more. An open discussion would not have skipped the acrimonious debate (we're still talking about Debian here), but it may well have led to something very close to what the release team is aiming for with strong buy-in from the development community. What the project will decide to do now is rather less clear; what we may be seeing here is the loss of a great opportunity.

The LWN site is currently under high scraper load, so comment display has been suppressed for anonymous users. If you are a human, you may read the comments by clicking the button below:

Note: you can avoid this step in the future by logging into your LWN account.