FOSDEM - Free and Open Source Developers' European Meeting, 26-27 February 2005, Brussels, Belgium

by Stuart Yeates on 1 March 2005 , last updated 23 September 2008

Introduction

I had heard about FOSDEM, naturally, but this was to be my first visit to Europe’s premier open source developers meeting. In the past year I have been to O’Reilly’s Portland based OSCon and Linux World London. The difference between those events and this one started at registration—FOSDEM literally didn’t have one. Sure, you could sign up for announcements on the website or make an (entirely voluntary) donation towards keeping it free, but no serious attempt was made to track how many people turned up and who they were.

The audience

It must take a great deal of confidence to book a 1470 seat theatre, a large auditorium, and an entire block of lecture theatres for parallel streams when you’re not even collecting registration numbers. Confidence…or experience. FOSDEM knows its audience and the FOSDEM audience clearly is committed to FOSDEM. By the end of the opening session it was getting pretty full. Not every seat perhaps, but a respectable attendance, especially given the fact that all the stands were manned at the same time.

The target audience of FOSDEM also differs from OSCon or Linux World. OSCon broadly targets everyone involved in Open Source. Linux World targets end users (and potential end users) of open source. FOSDEM solely targets open source developers - the heart of the open source movement. Every speaker on the main programme had heavy-weight developer credentials. With the exception of a small number now promoted into managerial roles, most remain developers first and foremost.

I was expecting FOSDEM to resemble the Portland meeting. But there appears to be a difference in open source culture on the two sides of the Atlantic. Portland was about business: most of the presentations in the main and side programmes were given by developers who were employed, full time, to do their open source development and used the buzzwords of the day to describe their work. FOSDEM, by contrast, was scheduled for a weekend because in Europe most open source developers have day jobs related only partly to their open source development work. They also tend to describe their work in more staid language. Many, perhaps most, of the FOSDEM developers were students, including some sufficiently young to require chaperones. At FOSDEM there was no enterprise readiness of mission critical applications. There were simple to install and use, distributed and robust systems—and everyone was expected to know that beyond the hype these are exactly the same things. There was no discussion of marketing and very little discussion of branding (with the exception of the on-going Mozilla fiasco), but a great deal of discussion about building critical mass and internationalisation—and everyone knew that to a compiler they are identical.

FOSDEM is also much more openly political than OSCon. FSFEurope and FFII were pride-of-place exhibitors. Richard Stallman spoke on political issues (and called for free BIOS); and the software patents issue was widely discussed.

Minority communities

Open source as a tool for the empowerment of minority communities (particularly minority language communities) was also very much the rage. There were success stories being discussed such as Welsh and Catalan. The Gnome and KDE accessibility efforts were also hailed, principally because they are engineered in such a way to provide necessary functionality and tools which work with so many of the projects developers work on while having so little negative impact on their development.

Key signings

FOSDEM hosted a number key signings over the weekend. This is as close as open source comes to a formal ritual. Key signings are curious gatherings at which developers put aside their much heralded meritocracy and become citizens. Each furnishes two forms of government issued photo-ID in order to prove who they are and give their public key. After the event participants use their own keys to sign the other participants’ public keys. The end result is that participants can not only communicate securely and free from prying eyes (which is relatively easy using https and ssh) but they can also know absolutely who they are communicating with (which remains an on-going problem when communicating on the Internet). Perhaps more important it enables developers to sign software, testifying that it came from them and has not been modified or tampered with. Almost all meetings of open source developers and users, from the lowliest LUG (Linux Users Group) to the largest conference, hold key signings, resulting in a web of trust enabling any open source user to validate the communication and signatures of any developer across the globe. International events such as FOSDEM and OSCon are especially importan since they gather together a diverse and international group of developers and facilitate opportunities for key signings.

The presenters

FOSDEM’s opening speaker was the infamous Jimmy Wales, who founded Nupedia and then Wikipedia. Wikipedia, the second incarnation of the Internet encyclopaedia, uses a bazaar model rather than a cathedral model, and recently hit the 130 million words and 482000 article marks. Wikipedia continues to grow at a staggering rate, with usage (bandwidth/page views) doubling every 3 months and several of the language collections seeing content growth rates in the range of 10%–20% per month. Jimmy sees developer effort lagging, to the extent that they have now had to hire a part-time hardware person, mainly to deal with failed hard disks in the Tampa server farm. Some of the Wikipedia spin-offs, such as Wikispecies and Wiktionary use more structured metadata, a significant departure from the free text used by traditional wikis. Considerable development effort is being put into finding the right representation of the structured metadata. In almost all cases the content editors know the functionality they want, but finding a generalised representation, across the sub projects, across cultures and across languages, that fits with the grass-roots wiki approach, is proving challenging.

Jimmy spoke about the internal dynamics of the Wikipedia community. He stressed that Wikipedia should not be viewed as the work million ants, each contributing a sentence, (as proponents of emergent theories claim), but rather the product of a community of thoughtful users. There are only a few hundred core editors to the English language wiki and only a handful in each of the other main languages (Japanese, German, Dutch, etc). These editors each make more than a hundred edits a week and enjoy a remarkable homogeneity of view on the direction of Wikipedia. Whereas Linus Torvalds attempts to be a benevolent dictator, Jimmy says he tries to work Wikipedia as a Monarchy, having as little as possible to do with any of the day-to-day technical or editorial issues, but focusing on building teams of editors for each language with a consensus building approach.

In the Calibre room (the only room which appeared to feature non-developers) Gregorio Robles and Diego Barcela discussed the growth and trends of the Debian stable releases. Debian is the largest Linux distribution by a significant margin, and appears to be the largest integrated body of software ever created, although there are no unambiguous comparison methods. Using standard software engineering metrics they discovered that while the number of packages is growing rapidly and successful individual packages are growing larger rapidly, the average size of packages is not increasing because of the flood of new (and by implication small) packages being added. The number of Debian developers is not growing as fast as the number of packages, meaning that each maintainer must look after more packages, but this does not necessarily mean more work for developers, since projects such as the Linux Standard Base and other standardisation projects are working to standardise software, making it significantly easier for developers to package third party software. There are also many tasks within the Debian project other than packaging software and sub-projects which make packaging easier, such as automated testing and internal standardisation. The traditional software engineering methodology used by Robles and Barcela followed only counts source code, non-source code such as developer and end-user documentation, localisations and art work are overlooked, even in projects in which they are central.

Rishab Ghosh (whom OSS Watch regulars may remember from our Open Source: national frameworks event) presented his latest findings on the interactions between developers and code in the Linux source code. While he raised many very interesting issues about open source communities and documents a number of enlightening correlations, as even he admitted, there are problems determining the direction of causality in many of the correlations.

The KOffice—Desktop Integration and Workflow Automation presentation by Raphael Quinet presented the current state of the art in the office application world. KOffice is rapidly approaching compliance with the OASISOpenDocument format, which will allow complete interoperability with OpenOffice. The second half of the presentation was a tutorial for the creation of KOffice plugins. KOffice is certainly built using some cool technologies, but it’s not entirely clear whether the project is going to be able to successfully compete with OpenOffice for developer effort. OpenOffice has several advantages over KOffice: it has the support of a large company; it has high profile, it is multi-platform and because it is not tightly bound to the large KDE systems new versions of OpenOffice appear in distributions significantly faster than new versions of KOffice. On a brighter note, the DCOP framework is to be replaced by the DBUS framework in the next versions of KDE and KOffice, putting them on track to interoperate more cleanly with the competing Gnome suite.

The FOSDEM dinner was somewhat under subscribed, but a number of the larger open source projects appeared to have arranged their own social or hacking events.

The next morning, I spent some time in the jabber developers’ room where several of the developers were addressing an issue that had already been raised periodically on the developer mailing list, the problem of zombies in chat channels. The reason it remained unresolved was that while the symptoms were clear (users presence remaining in chat rooms long after the users had left), it was unclear where the bug was—everyone assumed that it was someone else’s problem. With a couple of developers in the same room it was clear that it was someone present’s problem, and a test case was set up to find the root of the problem. After half an hours work killing servers to see how other servers reacted, the root of the problem turned out to be a slightly incorrect response to an error condition that propagated to clients. Error conditions are a common source of trouble because they are rarely seen and there are huge combinations of ways things can go wrong. They’re also very hard to test for during unit testing and release testing, so these standard software engineering approaches to reducing bugs are ineffective in dealing with bugs in error handling. I also had some questions about the possibility of running board games over the jabber network. My initial questions were invariably met with instructions to look at this document or that document on www.jabber.org, but after I’d learnt enough to ask meaningful questions I got meaningful, and useful, answers.

Alex Larsson’s Nautilus presentation focused on how to write plugins for Nautilus, the Gnome file manager. I’m sure the Gnome developers present were gripped by the material, but never having hacked on Gnome myself, I found it rather hard to grok.

Alasdair Kergon’s presentation on the Device Mapper detailed the on going work within the Linux kernel to provide advanced memory alternatives. Among other things the functionality enables admins replace or upgrade disks and take consistent backups without taking user services offline and is mainly required by those storing terabytes off data or requiring high availability. All the functionality is effectively hidden from users and common applications by the Linux file system. Most of the end-user visible functionality appears as LVM2.

Final thoughts

I greatly enjoyed FOSDEM and if you’re an open source developer, or looking to get into open source development, I highly recommend it.