linux.conf.au 2007: Friday 19th Jan

The final keynote, Kathy Sierra on Creating Passionate Users, was the first thing yesterday morning. Andrew and I had an interesting compare-and-contrast with Andy Tanenbaum’s Wednesday keynote when we talked last night. Sierra is very good, and her keynote was valuable in the sense of telling people things they probably sort of know in a really conscious, cohesive way so that they can make use of them. But it wasn’t as gritty in the way that Tanenbaum’s talk was, that is, it wasn’t something that most people both partially or fully disagreed with but also couldn’t shut up about. I didn’t hear anyone talk about Sierra’s talk with geeky passion.

That said, it was very good. Sierra argued (and in fact, writes and talks about everywhere, I’m told this talk was very similar to her OSCON one, and also see the Creating Passionate Users blog) that much as technical people don’t want to believe it, the human brain is not evolved to find protocol arcana or shutter speeds very interesting, because you can’t eat them or reproduce with them and you don’t need to run for your life from them. That’s not to say that everything about software needs to be presented in terms of food, sex or danger, but that the brain is evolved to pay attention to strong emotions and unusual images or behaviour, and that the emotions these inspire help people pay attention. If you don’t create a ‘twang’ of emotion, you won’t hold people’s attention.

There were two other points she made that I found interesting. One was about levelling as brought to you by computer games (and before that, pen and paper roleplaying). In particular she argues for regular partly symbolic rewards for progress to inspire users to continue learning and inspiring. The first two need to come very quickly, then they should come further and further apart. Ideally, the skill that you need to acquire to be certified at one level should be a skill that is integrally used the whole way through the next level.

The other was about outcome oriented documentation and teaching. This was the point that I didn’t feel people would really ever disagree with, but that doesn’t mean that it doesn’t have to be made. Her example was that of her camera manual. The camera manual is divided into sections like ‘shutter speed’, whereas the kind of documentation that Sierra was looking for was for tasks like ‘taking action shots [of her horses].’ This is the old user guide/reference manual distinction. It is easier said than done though, for a lot of things: you need to have a good idea of what these tasks look like for your tool to be able to write this kind of thing. That’s not always clear for software, especially for non-end user software.

After morning tea I was in the theatre for Dave Jones on Why userspace (still) sucks, but I wasn’t actually listening to it, sadly, I was updating a wiki page for the LinuxChix Blue Mountains trip tomorrow. It was a talk somewhat along the lines of the long ago Alan Cox post to nautilus-list on nautilus’s load time: applications polling the disk continuously and ludicrously for example, applications checking for a burnable CD constantly when you don’t have CD burner; applications that wake up and burn some CPU every ten milliseconds for the hell of it, that kind of thing. I’m told the subject of Ryan Lortie’s Burning CPU and battery on the GNOME desktop later in the afternoon was fairly similar.

I was bound and determined to get to André Pang on Concurrency and Erlang after missing his SLUG talk along the same lines. It wasn’t really long enough to demonstrate the superiority over threads with message passing instead of threads with shared state and locks except by passed along wisdom, but Erlang looks like a fun language.

I had a weird Jeremy Fitzhardinge experience at Open Day on Thursday: I read his wife Rachel Chalmers’ blog although I haven’t ever met and don’t know either Fitzhardinge or Chalmers; I originally used to read her Advogato diary and went from there. Anyway I saw a blond toddler wandering around at the Open Day and thought wha… why would a toddler look so familiar? It was Fitzhardinge’s and Chalmers’ daughter Claire, who I recognised from photos in Chalmers’ blog. But that wasn’t why I went to the Zachary Amsden, Fitzhardinge, Rusty Russell and Chris Wright talk on Writing an x86 hypervisor: all the cool kids are doing it! It was because I didn’t know what a hypervisor was. Well, it turns out that it’s the host operating system on any virtual machine setups, the one that handles traps, translates memory and IO syscalls and that kind of thing so that the guest machines don’t bite each other. Also, features kill puppies and everyone picks on Xen. So there you go.

The last talk I went to was Choosing and Tuning Linux File Systems by Valerie Henson. I’d already chosen ext3, so my work was nearly done. She recommends ext3 for most desktops and laptops; ext2 for filesystems where the redundancy is implemented on top and you want some speed underneath (apparently there are rumours that the Google File System is a layer on top of ext2) and XFS for heavy load with a lot of access (video servers).

I scoffed at us needing escorts to the dinner at the Australian Jockey Club, but actually it was a long hike under the racecourse. It was a difficult dinner in terms of its purpose. There were 600 or 700 people there, and most of us couldn’t really see the podium or the people at it; at the far tables they talked right through the speakers. The charity auction was over pretty quickly too; the bidding went into the thousands almost immediately, which rather stops things dead, given that that limits it to the small number of people who both knew there was an auction and have thousands of dollars. Our table had good fun with the attendee red and green glow sticks though, and ate lots of chocolate. I didn’t stay for the party upstairs and I’m glad I didn’t. I went home, ate fruit, drank water, and slept. And that’s all, folks.