User: Password:
Subscribe / Log in / New account Weekly Edition for February 26, 2004

FOSDEM 2004 trip report

Your editor was once told by a free software conference organizer that charging a registration fee was mandatory; without the fee, potential attendees would not take the event seriously and would not show up. The Free and Open Source Developers' European Meeting (FOSDEM) shows that this view does not always match up to reality. Perhaps uniquely among major Linux events, FOSDEM charges no registration fee, and, indeed, dispenses with the registration process altogether. That did not stop some 2000 people from showing up last weekend and packing the lecture halls to a level that would have sent a U.S. fire marshal into a complete panic. FOSDEM, clearly, is a successful event.

FOSDEM is organized in a way which is well described by its name: it is a meeting of developers. As such, it features a series of talks which are likely to be of interest to the development community and a distinct lack of presentations on how to configure the print system or on how Linux will leverage your business paradigm shifts into the next generation. Additionally, a set of "developer rooms" was occupied by various projects and interest groups (Debian, KDE, embedded Linux, Tcl, etc.). Each of those rooms was a place to gather, and most put up their own schedules of talks as well. Throw in a (problematic) wireless network, a beautiful city with no shortage of good food and beer, and support from a set of sponsors, and you have all the makings of a free software conference with a distinctly European flavor.

Keynote speaker Tim O'Reilly told the gathering that, while it is clear that free software is changing the computing industry, nobody, least of all the free software community, knows how. He pointed out that there are already user-friendly Linux-based desktop applications which are used by millions of people; they go by names like Google, Amazon, and Yahoo. These companies are building massive proprietary applications with free software, and, in many cases, giving little back. Tim would like to see free software developers think more about the use of their code in web application settings. He is also concerned about the implications of the large databases being created by these companies; those databases, too, are proprietary, and they can pose serious privacy threats. Do we, asks Tim, need a "web services bill of rights" which is analogous to the licenses which accompany free software?

Tim was immediately followed by Richard Stallman, who gave a fairly predictable talk about the importance of freedom, "Linux" and "GNU/Linux," etc. The freedom issues are important, but will be familiar to most readers of LWN. More amusing, perhaps, was the final part of the talk, where Richard addressed charges that he adopts a "holier than thou" attitude. Says Richard: "It's my job to be holy, I'm a saint." He then donned his disk platter halo and proclaimed himself to be Saint Ignucius of the Church of Emacs. Anybody can be a saint in this church, it seems; all that is required is (1) to free your computers of all proprietary software, and (2) make the profession of faith: "There is no operating system but GNU, and Linux is one of its kernels." (In the same humorous vein, Richard proclaimed that use of vi is not a sin according to the Church of Emacs; it is, instead, a penance).

Richard did also address the web services issue. He is not concerned about companies like Google failing to share their own code; what Google runs on its servers is its own business, and has nothing to do with anybody else's freedom. He is concerned about data stored on other people's servers; his response is to not keep his data there. Richard allowed as to how there could be freedom issues with web services, but he does not see those as free software issues in particular. One gets the impression he thinks he has taken on a big enough fight as it is; web services will be somebody else's problem.

There have been persistent rumors that a third revision of the General Public License would require that changes to code which are deployed in public web services be released. When questioned about this idea, Richard did not have much to say; there has been little time to work on such ideas, apparently, though that could change soon. He did mention the possibility of a "download source" clause. With this clause, the author of web-oriented software could include a "download source" link which would do exactly that. An optional license feature would require those deploying that code to retain the source download capability - and to ensure that it provides the source for the actual, deployed application. It is hard to see such an intrusive license winning a lot of followers.

The final keynote speaker was, inevitably, Jon 'maddog' Hall. Maddog talks resemble sitting in front of the fire with Grandpa and hearing his stories from before you were born. The stories are interesting, well told, and fun, but after a while you realize you've heard most of them before. You're always there when Grandpa tells another set of stories, however.

Keith Packard gave a heavily-attended talk on the future of the X server. In order to support many of the visually pleasing features envisioned for the future Linux desktop, some fundamental server changes will be required. In the new scheme, X clients no longer draw directly into the frame buffer; instead, they draw into off-screen memory which is then combined, under the control of a new "composition manager" process, into the screen seen by the users. Keith demonstrated some of his "eye candy" work which showed (1) how slick the Linux desktop can be, and (2) how slow it can be when all of this work is done in software.

In the future, Keith sees the X server moving into a fundamentally three-dimensional mode and speaking GL directly to the low-level graphics drivers. Many 3D applications will also be able to send GL directly to the hardware, and bypass the X server altogether. The current crop of two-dimensional applications will be handled in a compatibility mode. This change would pave the way for a new generation of 3D Linux applications, improve performance greatly, and would make vendor support easier; most video card vendors stopped wanting to deal with 2D modes years ago.

Keith also addressed the political issues currently being faced by the X community; see Zonker's article (below) for more information on that side of things.

LWN editor Jonathan Corbet presented two talks at FOSDEM; the slides from those talks are now available. The first was a variant on the "2.6 kernel changes from the inside" talk which has been presented at other events. Making its debut at FOSDEM was "kobjects, ksets, and ktypes: the device model from the bottom up," a low-level technical tutorial on the glue which holds the 2.6 device model together.

Other presentations seen by your editor include Robert Love on providing better support for the Linux desktop in the kernel (it is a good thing some developers are finally seeing this support as an important priority), Bill Haneman showing the features of the GNOME Onscreen Keyboard, Hans Reiser on the underpinnings of the Reiser4 filesystem, and an interesting developer room session on hacking into embedded Linux systems. There was far more going on than any one person could possibly see; FOSDEM is an event which truly showcases the vitality of the free software development community. It is not surprising that attendance has been growing strongly every year; this is one event which may have to find a larger venue for 2005.

Comments (9 posted)

X11: Where do we go from here?

February 25, 2004

This article was contributed by Joe 'Zonker' Brockmeier.

The XFree86 license change announced by the XFree86 project has caused a great deal of fuss in the development community. One month later, the new shape of things is beginning to come into focus. Unless something happens in the near future, the XFree86 Project's time as the custodian of the X Window System has come to an end, but X development will continue in a new home.

Ostensibly, the new license was to be applied as of the third XFree86 4.4.0 release candidate, but, according to longtime X developer Keith Packard, project leader David Dawes first checked in code under the license last September and updated the list of XFree86 licenses to include the license without any prior notice. Then the announcement that the new license was to be the "official" license for code copyrighted by the XFree86 Project was made by David Dawes at the end of January. The new license does not affect all code distributed by XFree86, but it touches enough code to create a major backlash among vendors and projects that are using and distributing XFree86.

The new license is a valid open source license, but it is a BSD-style license with an "advertising clause" that many find objectionable. The license is not GPL-compatible, which some say is a sure way to make a project irrelevant. Criticism of the new license is not limited to advocates of the GPL, however. It also seems to offend some ardent supporters of the BSD license, including Theo de Raadt:

Like other projects, we will not be incorporating new code from David Dawes into the XFree86 codebase used in OpenBSD. All such changes have to be skipped, rewritten, or you can contact the XFree86 group and place your own efforts to repair this damage.

This leaves the community at an impasse. With XFree86 sticking to the new license, and a large number of projects rejecting said license, other solutions must be sought. In the short term, many projects and vendors are planning on shipping XFree86 4.3 rather than using 4.4. Frederic Lepied, CTO of MandrakeSoft, says that Mandrake has reverted to XFree86 4.3 for the short term. Joseph Eckert, VP of corporate communications for SUSE, also confirms that SUSE will not be utilizing code licensed under the XFree86 1.1 license.

However, utilizing an older version of XFree86 is not a long-term solution. Daniel Stone, a Debian Developer, is one of many predicting a fork of the project to solve the long-term issues:

More than ever before, XFree86 has backed itself into a hole. The challenge now lies with the community to dig X out of the hole it's now in. Unfortunately, as kdrive and other solutions are not yet mature enough, it is my firm belief that this will only come about through a fork of XFree86. Sad, especially when you consider that that's how XFree86 came about; X.Org relicensed X, XFree86 got upset, and forked. We may be about to watch just a little bit of history repeating.

Keith Packard made it clear at FOSDEM that he believes this fork has already taken place; it was done by David Dawes when he changed the license. So now the "trunk" development effort is moving to According to Packard: and various Linux vendors are busy putting together a copy of the XFree86 sources from before the license change and are planning on making that available for developers to work on in producing X releases in the traditional fashion -- a monolithic release of the entire tree. The goal of this process is to ensure continuity of the window system implementation and allow people to get an X server capable of supporting more recent hardware.

Packard also says that the folks are working on improvements to the X architecture:

A related project that we're also working on is to take the monolithic X build architecture and splitting it into pieces. Libraries, fonts, servers and applications will be released separately. Periodically, released versions of the individual packages will be collected together and bundled as a unified release. The goal is to promote rapid development of some portions of the system (like video drivers) without requiring a rapid release schedule for the entire project.

As Stone said, we may be watching history repeat itself. Barring a change of heart on behalf of the XFree86 Project, it seems that projects and vendors making use of XFree86 will be looking elsewhere. The question is whether or not vendors will unify behind an X Window System produced by, or another group -- or if the fork ends up creating several splinter projects. With and several of the key developers behind it, looks well placed to become the new home of X development.

Comments (31 posted)

On Orkut

Some weeks ago, your editor was invited to join the Orkut service. Having never played with a "friend of a friend" service before, your editor found the experience to be naturally gratifying. After all, a system which inspires others to make public declarations of friendship cannot fail to delight such a stereotypical, socially challenged, geekish sort of person. It's nice to know that somebody likes you after all, even if you can never aspire to the triple-digit circles of friends that the truly cool people have.

That said, the free software community may want to think before committing too much to services like Orkut. A good look at the Orkut terms of service would be a place to start. It includes some relatively interesting things, such as prohibitions on reverse engineering and even (surprising, for a Google-affiliated site) indexing the site. The truly fun language, however, is:

By submitting, posting or displaying any Materials on or through the service, you automatically grant to us a worldwide, non-exclusive, sublicenseable, transferable, royalty-free, perpetual, irrevocable right to copy, distribute, create derivative works of, publicly perform and display such Materials.

So this site which, among other things, is supposed to facilitate business networking claims the right to make use of any idea which any user might post there. These terms may seem familiar: Microsoft attempted to get Passport users to agree to something similar three years ago. The company backed down after a public outcry; so far, however, Orkut users have been rather more accommodating.

There is a more fundamental question to be asked, however: if we, as a community, really want to document our associations, interests, sexual orientation, editor preferences, etc., do we really want to do so in somebody else's proprietary database? Social networks seem like a field in need of a great deal of experimentation; few people would claim that the best ways to aggregate, represent, and work with such data have already been worked out. If we're going to create a social network database, we should be doing so in a public manner that will allow free software hackers to play around with interesting new applications. We would almost certainly be surprised at what they would come up with.

One effort worth looking at is the FOAF Project. Rather than create a central, proprietary, indexing-prohibited database, this project is pushing for a distributed database built on individual RDF files. Such a scheme puts each participant in charge of their own data while making the whole network available for those who would create interesting interfaces to it. This project shows one approach to the creation of social network databases which avoids the problems of proprietary databases and restrictive terms of use. Doubtless there are others out there as well. We, as a community, do not need to put our time into the creation of somebody else's proprietary database; we can do better than that.

Comments (4 posted)

Page editor: Jonathan Corbet

Inside this week's Weekly Edition

  • Security: The trouble with backporting fixes; new vulnerabilities in hsftp, lbreakout2, and synaesthesia.
  • Kernel: <tt>invalidate_mmap_range()</tt>, ia32e and Linux; CDROM partitions.
  • Distributions: Linux in Brazil; Conectiva Linux 10 Beta 1; cAos 1.0; ekkoBSD BETA 2; OpenPKG 2.0: Reviews of Lycoris, ALT Linux, Slackware
  • Development: The Next Generation of Mail Clients, new versions of Firebird DB, Foomatic, AFPL Ghostscript, gnuplot, gDesklets, LyX, wxWidgets, GQview, Wine, Gungirl, Opie sdk, Epiphany, Aspell, Leo.
  • Press: Let Java Go, EU Council and software patents, MandrakeSoft loses in France, GIMP 2.0 review, openMosix intro.
  • Announcements: BizDev, Wind River and Red Hat, Wikipedia hits .5M articles, Lessig at OSBC 2004, LinuxUser &amp; Developer Expo, GCC summit CFP.
Next page: Security>>

Copyright © 2004, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds