FOSDEM 2009 report

Blog post by mmu_man on Mon, 2009-02-09 15:01

On Thursday I bothered the local print shop to get me some business cards to hand out, brought my own paper, cut them manually and I even didn't have enough money to pay him, but it was funny.

As last year, Rémi Grumeau, the BeOS France webmaster, kindly welcomed me at his apartment. I found out when booking train tickets that it was much cheaper to go on Thursday than Friday, so I spent the Friday in Lille, coding and walking in the town center. I also made a list of the talks I wanted to see.

Saturday

As last year, we got lost quite a bit between Lille and the ULB, and had to ask our way twice, since road signs density in Belgium is lower than that of BeOS users. We finally made it for lunch time and so did Olivier Coursière. We were soon joined by Alexandre Deckner who demoed Pulse on his dual core laptop. Niels Reedijk couldn't attend this time, hopefully next year.

Next to us we had the ReactOS project which I invited to join us for FOSDEM. Sadly it seems no one from Syllable I invited made it, but they really had their place secured. Maybe next year. The ReactOS table had many stickers, CDs, and they even brought a contribution box to put money in. That was a nice idea we could probably use as well.

We had a visit from Eric Bachard, one of the OSX porters of OpenOffice.org, whom I met at the last RMLL. He's eager to get an installable version with gcc4 to start porting.

Ithamar Adema popped up at the booth, with his associate. If you're into embedded stuff and need help, go check his start up. Hopefully he'll still find time to help start the ARM port soon.

I went to see Rob Savoye talk about his work on reverse engineering Adobe's RTMP streaming protocol in a much crowded amphitheater. It was quite funny. Rob is the author of the Gnash Flash player, he stopped by our stand last year to talk extensively about the Haiku port.

Måns Rullgård of the FFmpeg team stopped by to see Haiku. We often have trolls online about missing standard features in BeOS and porting applications to OSes versus porting OSes to applications, but he promised not to kill me.

Then I went to see talks about Java. The first one was about cross-compiling OpenJDK, for embedded ARM targets. Likely this work can be useful to ease bootstrapping the Haiku Java port. The other one was about the Caciocavallo framework, which essentially makes it much easier to write GUI backends for Java, by delegating AWT drawing back to Swing, leaving only the base canvas and top-level elements to implement natively. There is even a simpler interface that can use a dumb frame buffer and handles window managing from Swing. So one can implement native GUI support gradually by starting with a single surface with just 5 methods to write, then implement the higher level interface, and eventually the full AWT backend later on. This sounds like a good plan I think.

I heard other people tried to see me while I was away. Sorry folks, maybe next time!

We again got lost on the way back to Lille, as we didn't have a GPS or a printed map. We promised to print one as soon as arrived for the next day. Maybe we'll also think about using the train next year instead.

Sunday

On Sunday morning thanks to the map we arrived almost in time for Eric Bachard's presentation as he started a bit late too. Eric talked about how he implemented Apple Remote support in OOo Impress, detailing the OOo source code architecture. Then on another two hours Eike Rathke described debugging methods he uses on OOo on a real sample case. All this provided a first insight on the organization of the huge OpenOffice.org source tree, before launching the Haiku port for real.

Then after some not-so-dietetic meal I went back to the booth before visiting all other stands. I stopped by the XMPP (Jabber) stand to talk about the IM Kit. Another table I looked at had various embedded boards and PDAs, including a Zaurus 6000L and a tiny ARM based BeagleBoard. Other booths like OpenSolaris and Postgres were around. In the other building many GNU/Linux distributions had their tables aligned, along with Gnome and KDE ones. BSDs also had a good place. I had a chat with Rob Savoye in the stairs about how much Flash is not the Web, how much he dislikes Flash but had to come up with Gnash, and the Haiku port which gcc4 will likely get going soon. I talked a bit at the Firefox table, first in English then in French when we found out we all were from the hexagon.

At 3pm I went to see Theodore Ts'o talk about the new features in ext4, the biggest being extents, which BFS has been using for a decade as block_runs, and XFS as well on Linux. But it's always nice to see the main Linux file system catch up on that. Of course ext4 has many more feature BFS doesn't have, but this one was much needed to handle big media files, and with the new block allocator it turns out most files actually stay within 3 extents. After the talk a small group gathered around him outside the room to talk about various file system problems. Like the so-called "dot file" issue, which forces file system designers to counter bad behavior of applications regarding semantics on fsync() and ordering of writes to the file systems, which in case of a system crash ends up corrupting lots of those small files Gnome and KDE applications put in their respective dot directories… Then it went on power saving modes and fsync() abuse. Some others mentioned the lack of common method of determining the geometry of RAID arrays from the file system to optimize allocations. Then I wanted to ask him if he did any work on extended attributes in ext4 but barely had time to scratch the surface as he had to go for an interview. I'll probably mail him later I suppose.

Then I went back to the booth where it was almost time to pack up, which we did. We moved the tables back to storage, and went on to see the closing conference about how Google manages the Summer of Code and what's up for this year. Then I went back to Lille with Rémi almost without bad routing.

On Monday the sliding handle of my luggage that already got broken at BeGeistert completely fell apart when walking towards the station. I managed to put it back in and use the other handle and still made it in time. Luckily the return ticket was cheaper as first class so I had a power socket for the laptop to write this on board.

Conclusion

This edition of FOSDEM was really interesting, I managed to see more than one conference this time yet still not all I planned, again chat with many people on interesting topics, yet see a lot of people at our booth, despite it being located on a less crowded building. Actually it made it easier to discuss things than last year as the corridor in the main building is always full of geeks.

I believe the projects and people I saw at talks might really prove useful to speed up porting must-haves to Haiku, like Java and OpenOffice.org. I swapped cards with most of them, so we'll stay in contact.

Also, next year is the 10th anniversary for FOSDEM, and we'll likely have an R1/alpha with all the goodies (Firefox3, OOo, Webkit, Java, Gnash…) and the baddies (Java, Gnash… :-D) to show, so it promises to be even bigger.