July 25th, 2007
Nat Torkington, program chair, started off the day. He commented that one of the most interesting trends these days is the expansion of the Open Source ideals beyond software.
Tim O’Reilly commented about the FSF’s four freedoms, and asked how we maintain them. We have to think about preserving freedoms — questions such as Free Software that relies on proprietary services, data, or business processes. It’s important to remember to pay attention to freedom and not just to the success of businesses. But businesses matter and have enormous power and will always be related.
Tim really pushed expanding the boundaries of Open Source and thinking ahead: wikipedia, OpenID, etc. He also asked: does Congress need a version control system?
He suggested there are four open source success factors: frictionless software distribution, collaborative development, freedom to build/adapt/extend, freedom to fork.
Hadoop is an interesting FLOSS project to build some infrastructure like Google has. Apparently Yahoo is very interested.
Back to Nat… hardware is cheap and everyone keeps buying more of it.
James Reinders from Intel talking about multi-core parallelism. Saying that parallelism is going to be more and more important. Intel released threading building blocks, a series of templates for C++, as GPL’d software at the conference this week. I’m not all that excited about a C++ project, though, since I think languages like Haskell have more promise here anyway.
The other Intel guy mentioned Intel’s open source involvement: intellinuxgraphics.org, intellinuxwireless.org, linuxpowertop.org, kernel,org, moblin.org. Linux laptops have the longest runtimes compared to other laptops.
“It’s amazing how many people you can make paranoid by showing up with a tie and a suit to do a keynote at OSCON.” — James Reinders
Simon Peyton-Jones is up now, and Nat says he will “stretch your brain until only tiny bits are left.”
State of the art in parallelism is really 30 years old with locks and condition variables — like building a skyscraper out of bananas.
Locks are difficult to do right and have “diabolical error recovery”.
Let’s do transactions against memory instead of against a database. Implementation can even be similar to databases. The idea is transactional memory, and it sounds very, very slick.
Mark Shuttleworth and Tim for an interview…
Mark was fine, but I wish Tim had more interesting questions for him.
I went up to the front a few minutes after the event to talk to Simon PJ. He was talking to someone, who saw my nametag, and said, “Hi John, nice to meet you.” He looked familiar but I couldn’t quite place him, so I asked who he was. “Mark Shuttleworth.” Yep, I was sitting just far enough back from the stage that I wasn’t behind one of the large TV screens and couldn’t make out faces real well, and I didn’t recognize him. Erg..