Company: A Response to the Product Quality Online Petition

ShiyuTang

François Bancilhon, Mandrakesoft CEO, answers questions raised in a recent petition.

Hi,

Mandrakesoft values user feedback above all. Although some of you may feel your voice often goes unheard, please be assured that a lot of us spend a good part of our time listening to what users have to say. We do this for two reasons: the first reason is that the community is the very soul of Mandrakelinux – this is true of all free software. The second reason is that a company cannot live if it doesn't provide what its customers need – and this is true of every company, small and large.

So, although some emails may not receive a reply, and although we cannot intervene in every forum thread, we keep an eye on what's going on in the community. By signing this petition, you have given us a clear sign of your concern regarding the quality of our products, but more importantly you have shown us that you care about the future of Mandrakelinux. We share those two concerns. It is of the utmost importance for the future of Mandrakesoft and Mandrakelinux that the quality of our products be as good as we can make it, and that we do not disappoint our users.

Two specific issues appear in your in your statement : that of development and testing cycles, and that of the legitimacy of the Community/Official model. Allow me to answer the two underlying questions successively.

Mandrakesoft products can have one of three different upgrade cycles. Mandrakelinux follows a 6-month cycle, our mobile products (Move and GlobeTrotter) have a longer cycle, and our professional products (Corporate Server, MNF, etc.) have yet a longer cycle. We did not choose those durations at random – we have fairly good reasons for the choices we have made.

The main reason a new version of Mandrakelinux and related products (Discovery, PowerPack, ...) appear every six months is because, as you all know well, free software changes very fast and we must follow suit. Mandrakelinux has made a name for itself by always keeping up with the latest technology and integrating the latest software before every other major distribution. This is something we intend to keep doing. You are absolutely right in pointing out that this may cause, in certain cases, a decrease in system stability. This is exactly why we chose to keep with KDE 3.2.3 for 10.1 instead of going with KDE 3.3. We did not have enough time to make sure that this new version functioned well enough, and deliberately decided not to include it. Note that some users complained about that choice.

At every level, a compromise has to be made between stability and novelty. Mandrakelinux's 6-month upgrade cycle represents a level of compromise we think is acceptable for this type of product. We do not expect people to run mission-critical server applications on a Mandrakelinux which was downloaded just following release. We have products specifically made for that, Corporate Server being one example, and those follow a much longer cycle. Where optimal stability is necessary, Corporate Server is a better choice than Discovery.

With the now not-so-new release scheme, we've introduced another level of stability/novelty compromise. “Community” versions of Mandrakelinux are geared towards users for whom having the latest version of software such-and-such is more important than having to put up with a few glitches. It is not, however, a beta version. We provide smooth updates to Official or to the next Community version.

“Official” versions of Mandrakelinux, on the other hand, ought to be as stable as the kind of cycle chosen allows. Due to the incredibly wide range of PC hardware, we cannot test every configuration there is and that is another reason user feedback is so essential. Community versions serve two purposes: they answer some users' needs, and feedback from those users helps to improve the Official release.

We believe this to be a good system. There is a lot of pressure on Mandrakesoft, coming from all sides. We have to keep our users happy, and we also have to earn revenue. This pressure explains why the Community/Official model has yet to work in a way that is entirely satisfactory. It is going to work better for 10.1 than it has for 10.0, and we hope that in 10.2 everything will go according to plan.

Some of you may have noticed that we have started to improve our quality processes, this includes the testzilla repository and procedures which have been introduced with the release of Mandrakelinux 10.0. These new methods have already proven their worth – we hope the improvements they bring will soon be felt by end-users.

It is crucial that users of Mandrakelinux, who feel concerned about quality, participate in beta/RC testing, and report any bugs they encounter with their own hardware. A long release cycle does not in itself guarantee quality – if you are not convinced, ask yourself if the quality of Microsoft products, for example, correlates with the long cycles. To quote Eric Raymond (in The Cathedral and the Bazaar), the golden rule for Free/Open Source software is to “release early, release often”. We depend largely on contributions from you, our users, for the quality of the final product.

One of the ideas we are looking at is to organize a community test team which would not only report bugs, but would find bug fixes and solutions that could considerably help the QA process for Mandrakelinux products. If you have any suggestions which you think could help us improve QA, be sure to let us know.

In the meantime, it does not appear, from the feedback we got from Community users, that any major stability issue has surfaced. We have every reason to think 10.1 Official will be a version of Mandrakelinux as good as any, and it is our dearest hope that this will be your opinion as well.

Kindest regards,

François Bançilhon, CEO