User: Password:
|
|
Subscribe / Log in / New account

Tanglu: bringing Debian testing to the masses

Please consider subscribing to LWN

Subscriptions are the lifeblood of LWN.net. If you appreciate this content and would like to see more of it, your subscription will help to ensure that LWN continues to thrive. Please visit this page to join up and keep LWN on the net.

By Nathan Willis
April 10, 2013

There are scores of Debian derivatives in the Linux distribution kaleidoscope already, and there will soon be one more. A group of active Debian developers has announced Tanglu, a new community distribution to be based on Debian testing. The goals include providing a more up-to-date platform for upstream developers while also offering users an opportunity to use "pure" versions of upstream applications and environments. But Tanglu is not content to simply be a repackaged version of testing; it also wants to augment the system with many of the niceties currently found only in commercially-backed Debian derivatives like Ubuntu.

Debian has long been one of the most popular Linux distributors, in part thanks to the stability offered by its release process: Debian's stable distribution is just that—stable and reliable. The flip side, however, is that Debian releases can be quite slow to arrive, particularly when compared to the regular release cycles of commercially-backed distributions (for whom one or more "stable" releases per calendar year are expected). As a result, packages in Debian stable can lag far behind their upstream releases, leaving end users behind the times, and making it more difficult for application developers to test their code on Debian systems.

More adventurous users may choose to run Debian testing (which is more up to date than stable), but when testing enters a freeze (as it is now, for example), it remains frozen for quite a while. Once the freeze starts, Debian developers cannot easily get new package updates into testing. That makes it harder to test new packages, which is detrimental for the upstream developers as well as inconvenient for users wanting the latest release. Debian unstable is always another option (albeit not one that will be dramatically more up to date), but as a practical matter, many users and developers alike who are after a more dependable-for-daily-use system simply resign themselves to the situation, and settle for running and testing code on a "Debian-like" distribution like Ubuntu.

Testing-plus

On March 14, Debian developer Matthias Klumpp announced the formation of Tanglu, a new Debian derivative targeted at overcoming this impasse. Tanglu will be based on Debian testing and will track Debian development, he said, but it will incorporate new upstream package releases. The project will commit to a six-month release cycle, and will target desktop systems.

As Klumpp described it in the announcement, Tanglu is meant to follow Debian quite closely. He said both that "the delta between Tanglu and Debian should be kept as minimal as possible," and that "Tanglu and Debian should be working well together in mixed environments, where you for example have Debian servers and multiple Tanglu desktops with the new software, targeted at desktop user." To make that happen, he envisions existing Debian developers and packagers contributing packages to Tanglu. During a Debian freeze, he said, they could upload new packages to Tanglu first, and subsequently move them to Debian once the freeze is over. To simplify the process of contributing, Tanglu will "sync privileges" for Debian contributors, allowing them to update Tanglu's repository for the Debian packages they maintain. And Klumpp is quick to reassure doubters that he and the other initial Tanglu developers have no desire to stop working on Debian itself; rather, he said, working on Tanglu should aid Debian development because it will provide a good testing ground for packages while Debian itself is in freeze.

But Klumpp also describes some important differences between Tanglu and Debian. First, the new distribution will strive to create a "polished desktop which 'just works'," via a few add-on utilities. As he explained on the Tanglu development mailing list, those utilities include the Ubiquity graphical system installer and an Ubuntu Software Center–like application installer based on AppStream (which Klumpp is one of the co-authors of). In the initial announcement, he also noted that the distribution plans to use Ubuntu's kernel and KDE packages (as well as other Ubuntu packages), which should reduce the workload placed on the team. The announcement also says that Tanglu will include proprietary firmware and it will allow the easy installation of some proprietary software packages (although they will not be installed by default). Both of those inclusions are not permitted in Debian itself, which is one of the reasons Tanglu was started as an outside effort.

At the same time, Klumpp also seems keen to make Tanglu different from Ubuntu in some key respects. About a week before the Tanglu announcement, he posted a blog entry hypothesizing a Debian-based distribution. In that post, he expressed discontent with some of Ubuntu's technical decisions, and with Canonical's control of the decision-making process. The Tanglu announcement says that the new distribution will be KDE-based at the outset, but that the goal is to provide a "pure" desktop experience without "fancy modifications" to GNOME or KDE. The distribution will be community-run, he said, with feature proposals for each release cycle determined by a vote; "similar to Fedora, but without FESCo."

Implementation details

Klumpp's announcement garnered quite a few supportive comments, although a few people expressed concern that striking out on a new distribution (with a small team) was not the best way forward. Of course, positive comments alone do not a distribution make, and there is considerable work ahead before Tanglu can make a release.

Klumpp posted a follow-up report on April 8, describing what progress the team had made so far. Klumpp himself has been busy setting up the Debian Archive Kit (DAK) to manage the Tanglu package repository, a process he described as "not exactly fun." The documentation is lacking, he reported, and there are a number of Debian internals that are hardcoded into DAK. However, he was able to get it working, and noted that the work on DAK is beneficial to upstream (and even to his own Debian contributions, as he plans to implement some DAK changes for Debian's DEP-11 proposal). The automated build system is nearly finished as well, the initial import of Debian packages is complete, and the Tanglu bugtracker is up and running.

The initial set of Tanglu metapackages has been published; they make up a base system that includes KDE 4 and GNOME 3. The configuration also uses systemd as the init system, which is a change from upstream Debian, but is required for GNOME 3. In the status report, Klumpp declared this release Tanglu 1.1, although the project is not publishing any installation images. Rather, interested users will need to install Debian and then manually upgrade it to Tanglu, a state which some might consider premature to label a "release" (or, at least, a release with a version number greater than 1). Still, it is undeniable that the nascent project has made progress. Whether it is able to sustain itself in the long run, time alone will tell.

Tanglu is certainly not the first distribution founded with the goal of overcoming Debian's lengthy release-when-ready development cycle. One might consider Ubuntu itself (along with other Debian derivatives) just such an effort. More recently, the Constantly Usable Testing (CUT) initiative that was set in motion in late 2010 has a similar goal in mind; it makes monthly snapshot releases based on Debian testing, and performs QA testing to ensure that they can serve as installable images. Klumpp said on the mailing list that he regards Tanglu as filling a role distinct from CUT's, which he described as a development tool. To be sure, the inclusion of proprietary firmware and several Ubuntu features (including the live CD installer and a "software center") do place Tanglu several steps closer to Ubuntu's end of the spectrum than CUT, but exactly how much closer (and how Tanglu compares to other Debian derivatives) remains to be seen. But Tanglu and CUT do share a number of common themes, namely: producing a usable distribution that closely tracks Debian testing, so that users and developers can reap the benefits of Debian's stability, but also enjoy more up-to-date packages than those found in Debian stable. The goal is certainly an enticing one, but then again, having the best of both worlds always is.


(Log in to post comments)

Tanglu: bringing Debian testing to the masses

Posted Apr 12, 2013 8:09 UTC (Fri) by tdz (subscriber, #58733) [Link]

Sounds interesting to me. Thanks LWN for covering this.

I use Debian unstable in production

Posted Apr 19, 2013 13:13 UTC (Fri) by cypherpunks (guest, #1288) [Link]

It's actually pretty darn stable. The main thing is to update it frequently, so the updates are small and easy to reverse if necessary.

I do take some care. I read the changelogs and watch the upgrades for anything major. Sometimes I put stuff on hold. I upgrade my desktop first, then a test server, then slowly work across the cluster.

Then I pull really current stuff I want (Firefox 20, gcc 4.8, samba 3.6.13, etc.) from experimental.

But overall it's worked great. I haven't used testing (much less stable) in forever.


Copyright © 2013, Eklektix, Inc.
This article may be redistributed under the terms of the Creative Commons CC BY-SA 4.0 license
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds