Next to describing his hobby, it was an interesting overview of how Debian works, that a lot of noise is generated by a small group, some facts about the Debian Social Contract and a bit of history. New for me was that Read Hat considered using .deb in the very beginning, but they stuck to .rpm because of the work they put into it. Hp seems to support Debian.
Are trying to make something like Elektra, but only targeted to openwrt, which is very unfortunately. It does not use autoconf and its very limited in its functionality, but in contrast to elektra it seems like there is quite a community around it and it is heavily used. The kdb-tool is called uci there. They made quite a different design that "uci set" actually does not change configuration, but a "uci commit". You could implement that on elektra too, but it is not done. They are trying to have a backend support (especially something like the daemon in elektra).
This was actually the highlight of the FOSDEM. On a really tiny embedded hardware with an ARM 180MHz they showed some running full-blown Webapplications, using AJAX. Even better, Wt can fall back to static pages when the browser does not support it. They even got forward/backward of the browser working, incredible. I just looked a bit at there very active mailinglist and it seems like they really hit the right decisions: using boost::signal, boost::asio and boost::thread they don't lack of the problems of Qt, but have the same spirit. I will certainly look deeper into it and consider rewriting my homepage in it (I must get away from php4 anyway).
The talks was deeply technical (no problem from my side on that) but also heavily require a deep knowledge of postgres internals, which I unfortunately don't have. But it was encouraging to try out something great! I wish Postgresql to achieve what was said: To get the best Database in the world.
First confused by the completely other articulation than I thought, the talk was quite interesting. They try to use a (subset of) XPATH to express which key they want instead of a normal path to the key. The tree augeas builds is quite strange and apparently just takes over filesystem semantics with no abstraction. The (x)pathes are also a bit too long, e.g.: /file/etc/hosts/1/ipaddr compared with system/hosts/name in elektra (1 means the first entry, as if it had any meaning).
Whats really great in augeas and the technology I would love to see in elektra is the tree transformation. I feel very sorry that augeas did not perceive their chance to write a global configuration system. In the end its just a way to support some files lying around in /etc on some Linux systems. No integration. No abstraction.
Wow, that was a really cool hardware hack. Wookey explained how the solar panel work, how he did the control, a bit about 1-wire and showed lots of pictures.
An introduction in three new X-Server Extensions, namely Time, Audio and Compress. Unfortunately I missed the talk about XRandR 1.3.
Explained some concepts about how programs can (and should be tested): Take some valid input data and flip around some bits: See if the software crashes.
I am a bit unhappy about autoconf, mainly the dependency tracking and that I am not aware of m4. But before I learn m4 (which is not afaik used by anything else for a good reason) I consider bjam and cmake. There the talk about cmake just comes at the right time. Very nice is that you can also test and package your software with CTest and CPack (which is of course also possible with autotools, but I am not sure of bjam). What it is not so nice that it is developed by kitware rather than as opensource project.
Related: Alexander Neundorf about Why the KDE project switched to CMake -- and how
FreeDroidRPG is about 6 hours of great fun with quests and many dialogs.
A very professional talk about Engines, Library, Middleware, what it is, how it is related.
MuseScore is best (and only) opensource WYSIWYG music score typesetter. It uses high quality fonts (which have problems on mac os x, but look as great as lilopond), is written in Qt, translated in 12 languages and even have midi import and export
While I am happy with JFS many people have problems with ext3, but here comes ext4 to save their day. It removes (or enlarges) many limitations like the file size or number of subdirectories. A big advantage of ext3 against JFS is that it is a community project with many developers around the world, while JFS in mainly IBM. Ext4 seems to have some problems with many small files on disk, but hopefully elektra will make this optional some day.
Afterwards there was a (very long) private discussion about fsync for laptop and databases.
The last talk was about the pragmatic protocol groupDAV, webDAV and so on. The advice is to first support the groupDAV protocol and afterwards webDAV and then the extensions. Kontact seems to implement it very well, but they may have thrown it away during kde4 development, what a shame!