Sat 2 Feb 2008

linux.conf.au ... part 2

Things that happened at the conference since I last wrote. I flew back from Melbourne last night, missing out on the Open Day today, so this is only notes from the final day and some overall impressions.

Also, the talks are gradually being encoded and put up on the conference website in both video and audio format. I gather there are plugins around to let Windows users play ogg-style video, although I don't have any experience with it. Not sure if it's a case of "not yet" or a blunder that only half of my "stand up and blather" session is available, but it's hardly the most interesting thing there and it's impressive how fast the videos can be made available.

(For those wondering, "most interesting" might be Bdale Garbee's rocket talk, or Jon Corbet's State of Linux talk, or Stormy Peters' keynote or possible one I haven't managed to watch yet.)

Thursday Evening

The Professional Networking Session on Thursday evening was a bit of a mixed bag, as usual, but it's always something I look forward to.. HP helped sponsor it again and this always allows the organisers to hire a nice venue and put on a good spread of snacks and drinks. No change this year and many thanks to HP for this sort of contribution. The "networking" portion of the event is something I enjoy. Not so much as a chance to build up professional contacts, but an opportunity to talk to people who are giving presentations or spend their working hours up to their necks in IT. It cuts down the normal conference crowd from 600-odd to maybe half that or a bit less (I don't know the actual numbers, but I'd guess 200 - 250 people showed up.

Amongst other things, I had a chance to talk to Val Henderson about some of the details behind her talk (and express my admiration for the bash-based fsck she'd written). Also got an opportunity to thank Dave Jones for his "life of a distro kernel maintainer" talk. He'd prefixed the talk by saying it was all a bit depressing and I can believe that — most of the feedback you get are problems from irate users — but how people manage the polishing of community-based projects is something we don't hear enough about. So I'm glad he followed through on the request to give such a talk.

Friday

Friday had lots of good talks and was the day where I was conflicted as to which talks to see. I ended up seeing Dave Airlie talk about the state of ongoing development of X drivers for Linux, Matthew Garrett on suspend to disk (and it's non-workiness), lguest64, and library API design.

Dave Airlie's talk was very interesting. He gave a very comprehensible presentation of the state of development of drivers, both 2D and 3D, for Intel, NVidia and AMD graphics chips. It would have been very easy for him to make it a complaints session, but it was professionally done, pulling no punches, but focusing on the good and the possible over the bad. Compared to some of the less clueful comments that appear on sites like lwn.net in this area, Dave's presentation was very illuminating. I was surprised (that I could understand it) and pleased to have seen it.

Matthew Garrett is somebody I've seen speak before and he gives a nice presentation. His topic was a little controversial this time and some of the kernel developers in the audience kept wanting to chip in with their own criticisms. I would have preferred they saved that for their own talk, personally. Audience participation is nice, but it was clear Matthew knew about the problem space he was talking about and I would have preferred to hear him speak than the audience. Still, it's a fine line there, since, for example, Matthew Wilcox's contribution was invited by Garrett and definitely added something (but Willy knows when to speak and when to shut up, too).

Lguest (pronounced "rusty-visor") is something I'm trying to understand as a way of pushing my knowledge of the kernel closer to the hardware. Unfortunately, my tutorial on Thursday clashed with Rusty's tutorial on lguest hacking, so I missed that opportunity. Glauber Costa's talk on the work he's been doing at Red Hat, creating a 64-bit version of lguest was a nice addition, though. Very small audience, due to a number of interesting talks up against it and the relatively technical nature, I guess. Glauber, who I didn't know of before this talk, presented very well and clearly. He assumed some serious level of knowledge about lguest and CPU-level programming, but I was able to mostly keep up and learnt quite a lot. It also sounds like merging his work with Rusty's main tree and the main kernel's 32- and 64-bit merge for the x86 architectures is making things nicer under the covers for lguest, too. All in all, I'm glad the organisers accepted a highly technical talk like this; it was well done.

Finally, off to Eric de Castro Lopo's talk on library API design. I was hoping that nothing he said would be too surprising, since I've done a reasonable amount of this and have had my own experiences as a user/developer writing code against a lot of libraries. It mostly was a case of confirming my prejudices, but I mostly wanted to see Eric present, which he does nicely. The most amusing part of the talk was sitting next to Carl Worth in the middle of the packed lecture theatre ... Eric kept using cairo as an example of excellent API design and it's primary author was clearly pleased (with good reason; cairo's API is very understandable for us peons).

For once, I went to the lightning talks this year (as did most people, since they were the only thing on) and they were generally excellent. Everybody stuck to their time allowance perfectly, spoke clearly and gave enough clues that if you were interested you could go further. Some talks were received more enthusiastically than others, but my personal favourite was Mary Gardiner's three minute, five slide presentation about how to get a talk accepted at linux.conf.au. Apparently it's a little tricky and Mary did an entertaining and serious trip through what might work and what is a harder sell. It was close to a model lightning talk for my money.

Summary Thoughts

Organising linux.conf.au is ridiculous amounts of hard work, so it's always tricky to find problems, when so much that could have gone wrong did not. On the whole, it was again a very well run event and I enjoyed my time there, even though it was for only the main three days.

I'm glad the organisers kept all the main conference stuff (keynotes, in particular) away from the mini-conference days so that the "formal" and "informal" portions of the conference (a.k.a reviewed and not) were separate. No major slight against the mini-conferences, but they're quite different from the main event. Putting the tutorials in the mornings of the first two days was a nice idea. Personally, as a tutorial speaker particularly, I miss the three hour tutorials we used to be able to do, but I realise they weren't universally popular and certainly reduced the time available for other things. That being a given, this year's split worked well. You could see the keynote and then either go to a tutorial or engage in the social side of the conference for a couple of hours until lunch.

Fitting linux.conf.au into any venue is getting harder and harder. Rusty told me there were some problems this year with hiring a room large enough for the keynote (seating for 600+) and it was a bit squashy in the room we had (with temporary seating). The talk rooms also weren't so large .They were normal university lecture rooms, but only one of them was a huge theater. Since conference attendees are sometimes quite bad at telling time (a.k.a. turning up at the start of a talk) and sometimes a talk will run five minutes over, giving those people less time to get to the next one, it seemed like there was a constant trickle of people entering late. In a crowded room, that annoys me. As a speaker, it's nice that the hallways are packed with people, but it can't be comfortable. Again, I don't want to say it's a real problem with the conference, but it is a "problem" in a small sense, because it can detract from the talks. Adding another stream would be expensive. Shrug It's hard.

The social side of this conference is a big part. There are a number of people I see exactly once a year, in January. And a lot of bug fixing, development design, and general educating can happen in small groups. The organisers generally had things set up pretty well this year. The Melbourne Uni student union was open for lunch supplies (lots of reasonable food), there was a coffee shop right near the registration desk (part of the uni, again) and there was some seating inside to get out of the sun when we wanted to. So I think they succeeded in that respect. Apparently a few other experiments were tried this year with regards to the conference dinner (which I skipped) and in some other areas that I didn't take part in, but the comments from people I talked to were that they worked pretty well. I didn't really understand how the hackfest worked this year, but that seems to be a perennial problem after a very promising start a few years ago. A shame, but a relatively minor one.

So, I'm happy for the organisers (and grateful for their one to two years of preparatory work). I had fun; other people seemed to have fun; I learnt a lot; I met new people and caught up with old friends. The location (near to the city) was great, and the venue was good. It was best Olympics eva!

Topics: software/linux, conferences/linux.conf.au