August 1, 2007

What I learned at OSCON

Author: Joe 'Zonker' Brockmeier

Another O'Reilly Open Source Convention (OSCON) has come and gone. In many ways, OSCON is summer camp for geeks. A bunch of geeks from different parts of the world gather together each year in Portland, Ore., to learn, socialize, network, and get away from the workaday doldrums of our regular lives. Like summer vacation, it's over far too soon, but we pack a lot in just a short time.

I only have one minor complaint about the OSCON facilities. The wireless network is a perfect example of why Web 2.0 apps aren't going to take over the world anytime soon. Several times early in the week, the wireless network crapped out, leaving attendees without Internet access until the network team could respond. Until you can attend an event like OSCON and have rock-solid network connectivity throughout, users are going to want client-side applications and locally stored data. Otherwise, OSCON was well-organized, as usual.

I was impressed with the quality of the talks I attended. OSCON does a good job of bringing the movers and shakers out to talk about their technology. For instance, Larry Wall and Guido van Rossum were both on hand to talk about the status of Perl 6 and Python 3000, respectively.

I didn't get to attend the Python 3000 talk, but I did sit in on Wall's update on Perl 6. The room was filled to capacity (and then some) by attendees interested in hearing about the status of the software. From the look of it, the language has undergone quite a few changes since last year's talk. Wall's talk covered only Perl 6 the language, and didn't encompass Parrot, which will be the virtual machine that actually runs Perl 6, as well as other languages. The date for a final release of Perl 6 is still not set.

Improving performance isn't easy

Josh Berkus, one of the PostgreSQL core team members, gave an informative talk called "Performance Whack-a-Mole" in which he discussed methods of improving performance for online and database-driven applications. Berkus said this was a shortened version of a tutorial he gave earlier in the week.

Berkus talked about the vagaries of performance tuning, and what applications and parts of the system one should focus on for the best effects when performance tuning. Berkus walked the audience through the high-level concepts -- at what system level performance problems could be found, which problems to tackle first, and so forth. For instance, Berkus says that "as a rule of thumb less than 10% of performance problems are going to cause 90% of your problems."

It may not be worth spending a lot of time and effort to track down some performance problems because the return -- the improvement in system performance -- might not be noticeable. On the other hand, some performance problems may be the bottleneck that causes a major hit to performance.

Berkus also said that "the biggest performance problem is going to mask any other performance problems you may have." In other words, until you solve the biggest performance problem, you may not notice other "moles" that are causing issues -- which also means that performance tuning may be a long process as you discover smaller and smaller performance moles.

Though this was the abridged version of the talk, Berkus gave the audience some good starting points to help in troubleshooting application performance issues.

It's still the community, stupid

The community theme was a huge part of Ubuntu Live earlier in the week, and that carried over into OSCON as well. This year several sessions were all about community, a topic that wasn't prevalent last year.

On Thursday I went to a panel session titled "The Art of Community," which featured Danese Cooper, Brian Behlendorf, Karl Fogel, Sulamita Garcia, Dawn Foster, Jimmy Wales, and William Hurley (who goes by "whurley"). The panel touched on things such as "what makes a healthy community?" (best answer, "no killings"), community horror stories, whether a community can get too big, whether communities can be bought when an open source company is sold, and how to motivate communities.

Each of those topics is fairly broad, and could have had a session in its own right. Each of the speakers could have given a talk on any one of the topics and probably have filled the session time, so trying to cover them all in the alloted 45 minutes was pretty much an impossible task. Instead of getting deep into any of the topics, the panelists only had the opportunity to offer one or two thoughts, and then it was time to move on. While entertaining, the session was fairly light on takeaway value.

Have fun and be inventive

I have rarely laughed as hard as I did on Friday during James Larsson's "Pimp My Garbage" session. While not strictly related to open source, Larsson's talk fit with another O'Reilly publishing area -- specifically, Make Magazine.

Larsson showed pictures and video of some particularly inventive and funny hardware hacks, including one that brought the house down -- Leather Fetish Pong. Larsson described taking one of the chips from an old Pong game and then wiring up two leather boots with sensors so fondling the boots would move the controllers in the Pong game.

As if that wasn't enough, Larsson then showed how he added a riding crop between the boot controllers so that when a player lost the ball, they would be "punished" with the riding crop. The audience rocked with laughter when Larsson showed video of the system in action.

Failure is important

Scott Berkun's talk, The Myths of Innovation," covered the history of popular innovation stories, and then moved on to reasons why we tend to think of innovation incorrectly. Berkun pointed out that history tends to focus on the last successful steps prior to innovation and downplays the events leading up to new inventions.

Berkun pointed out that "most of our stories [of invention] are about the tip of the iceberg" and ignore all the work that went into the invention beforehand. Failure, and the willingness to fail in pursuit of something, is just as important as success -- because it's rare to succeed without many preceding failures.

Another "myth" that Berkun tried to debunk is the idea that people like innovation. He pointed to the Luddites as a particularly vivid example of people objecting to innovation. Of course, this is specific to innovation and change that affect one's ability to make a living.

Berkun also demonstrated that, looking back on events, we tend to view inventions and innovation (as well as other events) as inevitable. That, of course, is ridiculous -- there's no irrepressible tide that will sweep us toward a specific future, and thus we need to work to get to where we want to wind up.

I think this is a particularly salient lesson for the FOSS community. Many dignitaries in the FOSS community like to speak about the successes of FOSS to date as part of a movement that will inevitably carry us toward a future with greater freedom, where FOSS will dominate and proprietary software will wane in importance and influence. That certainly may happen, but it is not guaranteed, and we only need to look to the past to see other movements that have gained popular support only to succumb to the status quo.


  • Events
Click Here!