Within Reason

EuroBSDcon 2005 Fallout

Posted by Ceri Davies Tue, 29 Nov 2005 23:18:00 GMT

EuroBSDcon 2005 was a resounding success for me. I had lost a large amount of enthusiasm for the project, which has probably shown in the contributions from myself and my dealings with others.

After the weekend in Basel, I can categorically state that BSD fucking rocks, and you’d do well to use and learn it now before everyone else realises and you’re just another guy who has no idea about the “great new OS that actually makes your life easier”. Straight up.

Adrian Steinmann's talk on his Single User Secure Shell project was very interesting; on top of the actual subject, I learnt a neat trick about using kenv(2) to pass boot-time hints to the system. Adrian mentioned that having crunchgen(1) support "mostly static" binaries would be very useful for him; I hacked together a patch on the bus home, which should be ready for general release and testing soon.

I had meant to attend Ryan McBride's talk on network stack randomness in OpenBSD but accidentally went to the other room where Marc Schiesser was giving his talk on complete hard disk encryption. This talk ran out of time a little due to technical difficulties and a number of people, including myself, had to leave before it was finished (and critically, before we could give a well deserved round of applause; sorry Marc). Marc pointed out that a boot-time GELI(8) mount cannot use a key file; I may try to fix that using kenv(2) pointers to the key files.

Next, Mike Silbersack gave his talk on network stack randomness and interoperability which raised an interesting interoperability problem between OpenBSD and FreeBSD or Windows XP SP2, which a couple of OpenBSD developers said they would take a look at.

New evolutions in the X Window System was full of eye-candy, but seemed to be taking a lot of ideas from Mac OS X (although it's possible that Apple just beat them to the implementation, I don't think that's true) and is actually slightly worrying. We need to be able to keep up with the Linux camp on DRI implementation or we'll fall behind pretty quickly, I think. Luckily, none of it actually looked useful. It's also unclear what this new stuff means for architectures other than PC ones.

Emmanuel Dreyfus' talk on remote user access VPNs was great and I learnt a lot about aspects of IPsec from it. I was a little disappointed to see that the client solution for Windows and Mac OS X was Cisco's VPN client, since it seems a little perverse to buy a Cisco VPN Gateway for the client software only! I'm pretty sure that there will be another Mac OS X client solution at least though. I'm also a little disappointed that this was one of only three NetBSD talks. For a project that does marketing so well, it didn't seem like much.

Ryan McBride's talk on OpenBSD and PF was absurdly good as usual, with the benefit of new code marrying trunking and CARP allowing the use of an axe for simulating link failure. I know that if I demo this at work, we will use it; need Soekris machines.

Given the absence of The Real Robert Watson combined with the insanely detailed paper that he wrote, I went to the Failover Mechanisms for Filtering Bridges talk by Massimiliano Stucchi next. This was pretty rushed and seemed to be over in 20 minutes, during which time I had also been paying too little attention; my conclusion that it looked a little too much of a hack may not be quite fair as a result.

On the Sunday, Ollivier Robert's talk on distributed version control systems was interesting, though a little too early for my liking. I don't really see the need for one myself, but if it makes life easier for others (and is Mercurial), I say go for it. It's just not really important to me.

André Opperman's pair of talks were very good. The first, covering new networking features in FreeBSD 6.0 was a nice summary of stuff I knew anyway, but the next on optimizations for FreeBSD 7.0 was a fascinating discussion of the challenges involved in pulling a packet off the wire and processing it fast enough in order to be able to do it 60 million times per second!

The final talk was from Ted Unangst on OpenBSD's new threading library, librthread, which looks really lightweight and functional. Ted mentioned that there were some issues running MySQL; when my chat with Greg Lehey later that evening revealed that MySQL were looking for a lightweight thread implementation it was nice to throw the pair together. Hopefully something good will come of it.

After the panel session and some great music at the closing session, it was off to the hotel for a sleep, before meeting up with some guys at the restaurant across the road, where Geoff Buckingham and I indulged in a "we hate MySQL" joint rant at Greg Lehey. He tried to disclaim all responsibility, of course :-)

I also got pegged by Marius Nuennerich for not committing his PR yet, after which we had a good chat about a whole bunch of stuff. Talking to David Malone the next morning over breakfast, he mentioned that someone had had a great idea for a PR BOF, at which we would all line up and users could come and have their PRs looked at. Theoretically there should be enough committers at the meeting to decide if the fix was correct there and then.

In summary, I had a nice time and now I love BSD again. Next year I will have to be a little less shy, get off the wall and introduce myself to more people.

Finally, Philip Paeps wants me to talk at FOSDEM in March; I can't think of anything I know well enough to talk about — I'm open to suggestion or general building up ;-)