DistroWatch Weekly

AW: Obviously it never helps the mood at a company when people leave, and a lot of the people who left recently had extensive internal connections and it's sad to see them go. But there is a sense of optimism around our latest projects and products (including Corporate Desktop 4 and new Flash variants), and around some of the changes happening with the Mandriva community and the development process.

AW: My personal thought is that there's a kind of lag time for Linux distributions. Our problems really started to set in around the release of 2005 LE (10.2), but this wasn't really immediately obvious from the outside: we still got the press time, we still had the user base, we still looked like everything was going great. By the same token, when you start to fix problems (both with the distribution itself and with areas like support and community involvement), you don't see the results right away - it takes longer for people to get over their bad memories of prior problems. You still see a lot of people criticising Mandriva for issues that were resolved several releases ago, or saying they won't try new releases because of the bad experience they had with 2005 or 2006. For us, this is obviously frustrating, but we have to accept that it's a consequence of mistakes we made in the past, and be patient.

AW: Mandriva has always had a culture of openness, where differences of opinion between staff members are not suppressed. I'm sure anyone who's worked for any tech company (or, really, any company!) will tell you that front line workers often take different positions and perspectives from management.

Again speaking personally, I believe the problems that have affected the popularity of Mandriva are really quite simple. The most fundamental one is simply quality of the product. Around the time of the 2005 and 2006 releases, the bar for Linux distributions in some senses was being rapidly set higher by various competing products, and Mandriva took some time to adjust. We still had our historic strong points - hardware compatibility, the Mandriva Control Center, user-friendliness, our extensive development community - but the improvements made by many of our competitors meant that it was no longer the case that our strong points were sufficient to make people overlook our weak points, which tended to be mostly in the area of quality control and overall polish. 2006, especially, was released with several clear bugs and deficiencies which absolutely should have blocked the release. We definitely learnt from that, and for both releases since we have had a process around release time which has ensured that all really critical known bugs have been fixed before release. This is why the release quality of the 2007 and 2007 Spring products is far higher than that of 2006.

There are some more technical issues related to distribution quality which we have also worked on. Prior to the merger with Conectiva, we had a fairly primitive build system which relied on packages being built and uploaded manually by developers and contributors and then retrospectively synchronised into a CVS repository for tracking. Since the Conectiva merger, we have adopted the Conectiva build system more or less entirely. This is a far more robust system, which involves developers and contributors submitting source packages directly to a Subversion repository, from where the final packages are built in a controlled environment by an automated build bot service. This significantly reduces the occurrence of several build-related problems, makes monitoring of many typical problems (such as packages requiring rebuilds) far easier, and has also allowed us, since the release of Mandriva Linux 2007, to implement a new repository system which makes it very simple to provide both security / bug-fix updates and also new version backport packages for stable releases. There have been several other improvements of this nature which will have a consistent and ongoing effect on the quality of the distribution, a trend which should become obvious after a few more releases.

Beyond this, our problems have been in the same areas as they long have been: problems with service reliability, a tendency to overreach in the development of new products and services (which are then often abandoned), some inefficiencies in providing customer and technical support, and communication problems. Mandriva has been a bit behind some of our competitors in terms of having a culture of communication, both internally and among our developer community. A lot of people just don't hear about a lot of the cool stuff and hard work that goes on in the Mandriva world, and this is something we've been trying to address for a year or so now, with things like Planet Mandriva, the Mandriva team blog, improved communication with news sites, and my own role within the Mandriva community.

AW: Still speaking personally, no, I don't think so. Both markets are equally attractive for Mandriva. The home desktop market does seem the most immediately attractive, but it comes with its own considerable problems. You mention the new darling Ubuntu; trying to sell a desktop Linux product that is competing against the free-as-in-beer Ubuntu, the free-as-in-beer openSUSE, various other popular free-as-in-beer distributions (including the extensively Mandriva-based PCLinuxOS) and even our own very capable Mandriva Linux Free edition presents its own unique challenges. I've been trying to promote the viewpoint inside the company for a while now that selling software, per se, in the current Linux market is not a very attractive position to be in, when so many of the competitors, for various reasons, do not need to sell software. It puts you at a natural disadvantage.

The home desktop market is an important one for Mandriva for a variety of reasons: it does still provide a significant proportion of our revenue, and it has special significance for the Mandriva project as it's where we started out and where many of our community members and contributors have come from. So for all these reasons, we will be maintaining our commitment to it. The enterprise market has been increasingly important to Mandriva for the last few years, and if you examine our financial results, you will see that we have actually been quite successful in expanding our business in this market. We have considerable cost advantages over our main competitors in this space for certain types of enterprise, and we also have the advantage of being the only truly European-based enterprise distributor, which is attractive to many European companies. I think it's a stronger position for Mandriva to continue to address both markets than to attempt to focus on one.

AW: We want to make sure it works perfectly! The original Flash did not come out until several months after the release of 2007, and the 2007 Spring Flash will be no different. We have to test the product carefully before release, as it's rather hard to fix any problems once we've committed to an image. We hope this won't dissuade anyone from buying a Flash, as we will be running an upgrade program for Flash 4GB owners to update their Flash to 2007 Spring, the details of which haven't been finalised yet.

AW: We also like having Flash come out a few months after the main distro from a sales / marketing point of view. The reason is that otherwise we have a very cyclical sales situation: every six months, big noise and whoopee and sales for two weeks, then five months and two weeks of basically sod all. Nothing new to promote, market, discuss. If we have the Flash on an offset cycle from the main distro, we have something new to promote, market and discuss every three months. It flattens out our sales curves a bit and gives us a chance to be in the news for more of the year.

AW: I see 2008 as a kind of 'new start' for Mandriva, while building on the improvements in quality that were seen in the 2007 series. It will feature a lot of significant user-visible changes. Most obviously we will be moving to a new kernel version, after staying with 2.6.17 for the last two releases. All the other new versions you'd expect to see will be there - GNOME 2.20, the latest Mozilla suite and OpenOffice.org, and so on. There will be some headline features along the lines of the Metisse support in 2007 Spring, of course, but we're not revealing these yet! Beyond this, though, there will be quite a lot of fundamental updates. Personally, I have been working on a project which is quite close to some of our community members' hearts - rebuilding all the old packages in the distribution which, for various reasons, have not been rebuilt for several releases. The new Bug Squad has been cleaning up the bug database, which has led to a pretty big rush of fixes for old bugs. In general, there's a sense of 'spring cleaning' in the development process at the moment. I expect the overall impression people get of 2008 will be one of improved reliability, maybe with things they thought might never get fixed or improved being addressed.

Our awesome development community is also coming up with some great stuff. In particular, a few of the development volunteers have formed a team to focus on improving Xfce within Mandriva, to make it a really top-notch desktop alternative alongside KDE and GNOME.

One personal project I'm hoping may bear some interesting fruit: I've packaged Gtk+ WebCore in the Mandriva development distro. Gtk+ WebCore is a GTK+ port of Apple's WebKit web rendering toolkit, which is of course in turn based on KDE's KHTML, and which is the basis for their Safari browser. Gtk+ WebCore was started a couple of years ago by Nokia, brought to a minimally-releasable state and then kind of left dormant. Recently, the lead developer has been active on it again, and there's been an initial release of a new Gtk+ WebCore based browser, Midori, which I've also packaged. Both Gtk+ WebCore and Midori are pre-alpha state and not really usable for serious web use yet (on the one hand pages actually render really well because the fundamental technology is very mature, but on the other hand the browser has almost none of the features you'd expect, and the whole thing crashes at the drop of a hat), but it's a fun technology to play with, and I'm hoping that by the time 2008 gets released it will have been developed to the point where you can at least use it regularly if you're really determined! It'll be nice to have a GTK+ alternative to the Gecko-based browsers.

AW: Near the start of the release cycle, we start both an internal discussion and a discussion among our development community to decide what features people are interested in working on for the next release. You can see some of the public side of this in this Wiki page (the rest you'll find in mailing list archives). There's a similar page on our private internal Wiki which covers the goals for each product team. Besides this, the open nature of Mandriva development means sometimes people can just turn up and do things. Our leading position in support for 3D desktops is quite heavily based on work provided by a contributor, Colin Guthrie, who just showed up during the 2007 development cycle and started submitting work to make the 3D desktop technologies work smoothly with Mandriva. With a little help from Olivier Blin, who is the internal developer responsible for drak3d, the whole project was born. We're occasionally accused of wasting too much time on 3D desktop stuff that we could be spending on 'more useful' work - people who make this criticism usually don't realise that 3D desktop work takes up only part of the time of a single one of our paid developers!

The other team with an input is the marketing team. They can suggest some features and ideas based on what they think will be attractive to users.

AW: No, Metisse won't go the way of Kat. Metisse is a more robust project with a proper team and funding behind it. Also, to put it simply, Metisse works a lot better already than Kat ever did. Kat was a great project, but it was really too immature to be included with Mandriva at the time it was, and that put a lot of pressure on the developers. Ultimately, it was overtaken by other technologies.

AW: Well, I hope you can see from the above comments that this is not the case at all! The Brazil office has actually become increasingly important to the development of Mandriva. For instance, right now, the maintainer of Firefox and Thunderbird is a Brazilian developer (Marcelo Leitner). He's also responsible for printing. Gustavo Boiko is our X.Org maintainer. Luiz Capitulino is in charge of kernel updates for stable releases, and he did an excellent job on the 2007 series. Helio Castro is heavily involved in KDE work: he built the KDE 3.5.7 Mandriva packages that are available at kde.org, for instance. Andreas Hasenack packages various important applications (including Postfix), and is heavily involved with developing and maintaining the build system. Claudio Matsuoka is our Bugzilla administrator. Really, the fact that you don't hear much about 'the Brazilian office' is an indicator of how successful the merger has been: we really don't often think in terms of a 'Brazilian team' and a 'French team', we just see one big development community. The Brazil-based developers work together extensively with both the French developers and the wider volunteer community.

AW: The most common misconception about Mandriva is that it's 'not free'. This takes several forms - the belief that you have to pay or join the Club to get updates, or that the freely available versions of Mandriva are crippled demos. In truth, we have one of the longest and best records of freedom. The Free edition of Mandriva, which is both free as in speech and free as in beer, has been going since Mandrake began in 1998. It's not crippled and it's not a demo: it's a fully-functional distribution which is used as a primary operating system by thousands of people. Every free / open source package in the Mandriva distribution is available to the public at no cost from our repositories, none of them are reserved. All official updates are, and have always been, available without payment, membership, subscription or registration to anything. We used to differentiate our commercial editions on the basis of the non-free software included in them, but in recent years, with the importance of non-free software to Linux users declining and the amount of non-free software we make available at no cost increasing, this difference has become very small, and we're increasingly focusing on the support, services and documentation that come along with a purchase of a commercial edition.

The best 'feature' of Mandriva that most people don't know about is the new repository system. Since the release of Mandriva Linux 2007, we have been using a new, segmented repository system. The best thing about this is that it allows us to provide an official, cohesive source for new versions of software for stable releases. The repositories for each Mandriva Linux release are split into four sections: main, contrib, non-free and restricted. Main contains free / open source software that is officially supported. Contrib contains free / open source software that is not officially supported. Non-free contains non-free software that is available to the general public, and restricted contains non-free software that is only available to purchasers and Club members (it's a very small section).

This split has always been in place, but what's new is the subdivision of each section into four repositories: /release , /updates , /testing and /backports . /release is the 'reference' repository: it contains the exact version of each package that was current at the moment the release was finalised. /updates contains security and bug-fix update packages - it's the repository used by MandrivaUpdate. /testing contains candidate updates: when a bug is reported and we plan to release an update for it, we first place the update in /testing for interested users to test and verify that the bug is fixed and no other problems are caused. /backports is the repository for updated versions of packages. We don't put these in /updates , as it's not good policy to mix updated versions with security and bug-fix updates, and users in conservative environments don't want this kind of update. However, many users do want to have new versions of major applications when they're released, and the difficulty of providing these has always been one of the biggest drawbacks of the standard distribution repository system of software installation in Linux. The /backports repositories solve this problem for Mandriva, making it trivial for packagers to provide updated versions of applications for stable releases, and easy for users to use these packages if they choose.

In Mandriva Linux 2007 Spring, using the /backports repositories is as easy as running the repository configuration tool, adding the standard set of online repositories, and then enabling the /backports ones (by default, they are added to the list, but not enabled). This system has been a huge success for 2007 and 2007 Spring, and it has allowed us to provide such important new releases as Pidgin, new versions of Mozilla Firefox, Ekiga, Xfce, SeaMonkey, vdr, freevo, Enlightenment 17 and many more all in one official, centralised place, using the same buildsystem as all the other official repositories. It even allows us to add interesting new applications during the life of a release: for instance, Colin Walters' experimental GNOME shell replacement, Hotwire, has been introduced to the 2007 Spring /contrib/backports repository since the release of 2007 Spring.

Compare this situation to the case with other distributions, where if you want to add interesting new applications and keep existing ones up-to-date, you wind up adding several dozen single-purpose repositories provided outside the official repository framework, which are frequently unaware of each others' existence. To my knowledge, we're the only stable release-based distribution with a framework like this in place.