Why collaboration and free software make sense in the enterprise

No Comments

The words "sharing" and "collaboration" don't exist in the lexicon of your typical for-profit corporation. Josselin Mouette was at the Gnome Asia Summit to show how free software can favourably tip the balance sheets. Josselin, the Gnome maintainer in Debian, also talks about CUT, the new addition to the Debian family of repositories.

Mayank Sharma: What's involved in packaging Gnome in Debian?

Josselin Mouette: The Gnome team has been very enthusiastic about packaging Gnome 3. We have a specific repository in Debian which is meant for not-yet-ready technologies. We have a very strict policy in Debian to only upload technology ready for stable release, even if it is in our unstable repository.

But we have been working on Gnome 3 pre-releases, and atleast four people including myself have been working on it and I'd say three-quarters of Gnome 3 is almost ready. So it will ofcourse require a lot of polishing to reach the Debian quality standards but I am confident that Gnome 3 in Debian will be great.

josselin1.JPG

MS: I am sure it will be. But considering Debian's release cycle, when will it be?

JM: Everybody talks about the Debian release cycle. But they are comparing it with the release cycles of Fedora, and Ubuntu, which are extremely short. Debian does a distribution which is meant for everyone and it is meant to last. It isn't technically possible to make a distribution in six months that will last.

Before release we spend, for example, four months of full freeze, and it is just enough to be able to have a completely stable and ready to use in various environments distribution. The target users are simply not the same. So you cannot compare a Debian stable release with a Fedora stable release. It is often said that the kind of software you find in Debian unstable has about the same stability as a Fedora stable release.

Which is why to address the need of the home users that want the greatest and coolest software we are introducing with Wheezy, a rolling release named CUT for Constant Usable Testing. So our goal is to have the testing distribution which is currently only there to prepare for a new release. It was not until now meant to be usable, it was only meant to be releaseable, which is very different. We now want to make it usable too.

So this implies a bit of change in the process. It means more work for the team that decided to do it, but it is a great opportunity because it will bring in a new class of users who until now had to use the unstable repositories, which while being very stable, ofcourse are not guaranteed to be installable and it wasn't guaranteed that your system won't break tomorrow, even though such a thing happens rarely. In 10 years, I have seen it happen only two or three times.

MS: So how many repositories do you have now?

JM: We have five repositories. Old-Stable which is the former stable release which we maintain for atleast one year after the stable release to give time for people to upgrade. You'd think in a Fedora world it is extremely long but in an enterprise distribution world it is actually a bit too short.

Then you have the stable distribution which has the full support, and the security support especially. The testing distribution which will include CUT, the unstable distribution which is where the development happens, and the experimental which is a repository and not a distribution. It is normally on top of unstable but you can cherry pick the experimental packages that you want to run. And these ones can really wreck havoc.

MS: Why is Gnome the default desktop on Debian and not KDE?

JM: We just happen to pick Gnome as the default desktop environment because it is a bit more suitable to new users, and it is lighter and runs on more hardware.

MS: And you use Debian inside EDF, right?

JM: Yes. We use Debian a lot. On HPC clusters and on scientific workstations. This is where Gnome has its place. The Debian choice is really important for us because it really is a universal operating system and we can use the exact same version on our desktops, on our servers, and even on our clusters. This is unique. No other operating system allows you to do that.

Debian is even better in the area of scientific computing because there are a lot of scientific computing packages already integrated into Debian. So we have much less work to do to integrate the software into the distribution because the work is already done by the community.

josselin2.JPG

MS: At the summit, you are talking about making money from free software. How do you do that?

JM: At EDF, we use a lot of free software to run our scientific computing infrastructure but we also contribute to the free software community. For example if we find a problem with Gnome we won't just patch it for us and the same goes for everything we use in Debian. We forward as much as possible to the community because we don't want to maintain a fork.

But much more importantly, EDF needs a lot of software to run very complex scientific simulations. EDF is not a software company. We are an energy company. Our revenue comes from our customers who use our electricity. So it is not reasonable to expect us to be good with marketing software or generating revenue out of software. Software is always a cost to EDF. Always.

So how do you reduce this cost? We have several big in-house software that had to be developed especially for the nuclear area where the nuclear authorities are very strict and you have to simulate a lot of things to prove that your plants are safe. And the software grew to very very huge sizes and they became a real burden to maintain. So we decided to completely open them and not only put a tarball on the website. Today several of our major formerly in-house software are driven by consortiums or steering committees which are external to EDF.

So although we still have a word in them, most of the development is not done in EDF anymore and other people from other companies from different areas will contribute to them. We benefit from this a lot because we don't have to involve as many developers as we used to, and the software is better tested, and it becomes a reference. What used to be EDF-specific software is now reference software for the whole domain of computer simulation and everyone in this area uses it.

You'd probably have never heard of such software but when you have software of one million lines of code and two pages of features, with this software you can simulate just about everything in mechanics and it has an incredible number of applications.

MS: EDF does very specialized kind of work. So do you think this model of community-development can be adopted by other industries as well?

JM: I think it applies to a lot of companies that are not software companies. A huge number of companies develop in-house software just like us where our industrial secrets are not in our software. We just want software that is studied and is the most proven, the most reviewed the better it is for us. We don't want to keep this secret. What we can't share is the designs we simulate with this software.

In today's world where you have to collaborate with other large groups to build very big projects, and the latest energy projects always involve two or three very large companies, it is much easier to work together when there is free software involved. Between such large companies free software reduces greatly the amount of intellectual property you have to share with them and it simplifies processes a lot.

Images: Shashank Sharma (under CC By-SA 2.0)

Comments are closed for this post