I managed to attend the Wednesday morning keynotes, even after a night of beer and “Apples to Apples” with Dan, Brad, Alasdair, and Josh. It’s amazing how much better a shower and a couple cups of coffee will make you feel. Anyway, since the wireless network barely works once everyone has gathered, I was provided the chance to actually pay attention to the keynotes.
It was nice that the real Nat Torkington finally decided to join us. Until this morning, we were plagued by an imposter wearing a plain, boring shirt. I’m sad to hear that Nat won’t be returning to OSCON next year. First Randal, then Damian, now Nat? I’m starting to wonder if this conference will be worth attending next year. Makes me wish I’d started attending back when OSCON was in San Diego (my own home town!).
After the usual round of announcements, Nat explained that this year OSCON has a session track dedicated to people. How to work with them, how to design software people actually want to use, that kind of thing.
Tim O’Reilly’s keynote is made up of his radar, otherwise known as “What’s On Tim’s Mind This Year.” In a nutshell, he is pondering the future of Open Source in a Web 2.0 world. These days, many new things are built on top of services. Services that may be free to use, but may not themselves be Open Source. Even if the code running the services were completely open, in so many cases it wouldn’t be useful (perhaps it requires massive compute power to run, like Google). He warned the Open Source community to keep an eye of the future of information. He perceives a race to own the information that will power the future.
Next up was a bit of a repeat of Monday night. James Reinders, Intel’s director of marketing for Open Source products, stood up to plug a new product. Actually, Threaded Building Blocks isn’t a new product, but it has been newly open sourced. Intel is dedicated to support it on as many processors and in as many compilers as possible. They really want to be the drivers of parallelization in the future.
Continuing the theme of parallelization, Simon Peyton Jones from Microsoft Research in Cambridge, England, continued told us about the downfalls of locking models in parallel programming. He talked about task parallelism and what he called the atomic locking model. Instead of writing complex software to manage parallelism, traditional sequential code is wrapped to protect it from the pitfalls of parallelism—without the need to change the sequential code. This is great stuff, and something I advocate daily: remove the infrastructure from the programmer. Let the programmer focus on the task at hand and doing that task correctly.
All this focus on parallelism excites me. At work we have a huge cluster of compute power, most of which is dedicated to running individual sequential applications. Even with all of the multi-core systems we’re adding to our data center, we end up treating each core as a distinct processor. It’s a win when it comes to compute density in the data center, but we’re stuck in the old paradigm. The applications we’re using aren’t taking advantage of the increased parallelism to do more work in the same amount of time.
After all the talk of the future and parallelism, Tim O’Reilly interviewed Mark Shuttleworth, the head of the Ubuntu project. I don’t use Ubuntu, nor do I pay much attention to it, so by extension, I didn’t pay much attention to the interview. However, Ubuntu gets a lot of buzz in the Open Source community. I use (and have been known to sometimes contribute to) Fedora. More and more, I see Fedora—and, by extention, Red Hat—looked down upon by some in the Open Source community. I’ve even started to lose interest in contributing to the project. Is it still the right distribution for me? Where did my excitement about the project go?
Finally, there was a question and answer panel with all of the keynote speakers. I zoned out around this time, so I don’t know what was asked or what the answers were.