OSCON 2010

It’s nearing the end of July, which means that OSCON has come and gone. Here are my observations on this year’s event.

Talks

As always, there are a huge number of talks at OSCON, and this year I found it particularly hard to choose between sessions, though in several cases, hallway track conversations ended up making those choices for me. There wasn’t a theme to my talk attending this year, because a lot of topics are relevant to the work that I am doing now.

I attended a talk about Face Recognition on the iPhone. Sadly this turned out to be very focused on setting up for and calling the OpenCV library and less about face recognition, UI, or integrating face recognition into applications. As I’ve written previously, I think that new interface modalities may be arriving along with new devices, so I was hoping for a bit more than I got.

Big Data is also a topic of interest for me, so I went to Hadoop, Pig, and Twitter, and Mahout: Mammoth Scale Machine Learning. Hadoop, Pig, and Mahout are all projects at Apache, and each of them have an important part to play in the emerging Big Data story. The sort of analytics that people will be using these technologies for are part of the reason that data is now the big area of concern when discussing lock in.

The open source guy in me likes the idea of WebM, but it looks to me like there’s quite a way to go before it will be replacing H.264. I was surprised that the speaker didn’t have a better answer than “our lawyers must have checked this when we acquired On2″. More than anything else, getting clarity on the patent provenance for VP8 is what would make me feel good about WebM.

Robert Lefkowitz (the r0ml), is always an entertaining and thought provoking speaker. His OSCON presentations are not to be missed. This year he gave two talks, and you can read some of my commentary in my twitter stream. Unfortunately, r0ml picked licensing as the topic of his second presentation, and his talk was interrupted by an ill-tempered and miffed free software enthusiast, thus proving r0ml’s earlier solution that open source conferences are really legal conferences.

I’ve been following / predicting the server side javascript space for several years now. One of the issues with that space is the whole event based programming model, which caused mortal Python programmers headaches when dealing with the Twisted Python framework. Erik Meijer’s group at Microsoft has been grabbing techniques from functional program to try to make the programming model a bit more sane. I had heard most of the content in his Reactive Extensions For JavaScript talk before, and I’m generally enthusiastic about the technology. The biggest problem that I have is that RxJS is not licensed under an open source license. At JSConf I was told that this is being worked on, so I dropped in for the second half of Erik’s talk hoping to hear an announcement about the licensing. It was OSCON after all, and the perfect place to make such an announcement, but no announcement was made. I hope that Microsoft won’t wait until next year’s OSCON to get this done.

This year there were two of the keynote presentations that made enough of an impression to write about. The first was Rob Pike’s keynote on Go, where he eloquently noted some of the problems with mainstream programming languages. There was no new information here, but I liked the approach that he took in his analysis. The second was Simon Wardley’s Situation Normal, Everything Must Change . Simon is an excellent presenter and full of insight. While his talk was ostensibly about cloud computing, I think it was a little deeper than that. His story about the cloud is the story about commoditization of technologies, and since one of the roles of open source is commoditization of technology, I felt that there was some nice insight there for those of us working in open source. Simon also discussed the mismatch between innovation and mature organizations, another issue that people in open source often run into. The video is already up on YouTube, so you can make your own assessment of the talk.

The size and breadth of OSCON gives it one of the richest hallway tracks of any conference. This year was no exception – the hallway track started on the train from Seattle to Portland, and extended all the way through the return train trip as well. I always look forward to these discussions and to connecting with old friends. One friend that I caught up with was Cliff Schmidt, now executive director of Literacy Bridge, which is working to make knowledge accessible to people in poor rural communities throughout the world via their Talking Book device. Cliff had one with him, and this was the first time I had seen one. Just in case you haven’t, here’s what they look like:

Dailyshoot 249

Languages

OSCON began as a language conference, and this year, there were two special events in that space, the Scala Summit and the Emerging Languages Camp.

I have followed the Scala community pretty closely, because they are quickly accumulating real-world experience with functional programming. There are lots of cool tricks that remind me of stuff that I played with when I was in graduate school, and there are lots of bright people. But there are some things that I find worrying. For example, one of the speakers was touting the fact that Scala’s type system is now Turing complete. If I’m using Scala, one reason is that I want my programs to type check at compile time. Having the type checker go off and fail to halt is not what I had in mind. I recognize that you’d have to write some gnarly type declarations for this to happen, but still.

This was the first year for Emerging Languages Camp, and from what I can tell it was a roaring success. I didn’t attend as many sessions as I would have liked. This was due to a combination of factors – other talks that I wanted to see, being the biggest. The other factor was that the first talk I attended was Rob Pike’s talk on Go, and the room was very full, which made it hard for me to concentrate (probably had more to do with me than the room). When I saw that all the talks were being recorded and that the video folks promised to have them up in 2-4 weeks, it made it seem less urgent to try to pop in and out and fight the crowd. Still, this is a sign of success, and I hope that the minimum, the Emerging Language Camp will be given a larger room next year. Part of me would like to see it be a completely separate event from OSCON, but that’s probably not realistic.

Of the talks that I was able to attend, I found the Caja and BitC talks to be the most relevant. Fixing Javascript to have security is important for both client and the burgeoning server applications of Javascript. I wish that I had seen the talk on Stratified Javascript, since concurrency is ever the hot topic these days. As far as BitC goes, we are well beyond the time when we should have had a safe systems programming language. As much as C has contributed to the world, we really need to move on.

What is OSCON for?

I had a few discussions along the lines of “What is OSCON for?”, and Tim Bray shared some thoughts in his OSCON wrap up. As I have written before, I think that open source has “won”, in the sense that we no longer need to prove that open source software is useful, or that the open source development process is viable. There are still questions about open source business models, but that’s a topic that I’m not as interested in these days. Open source having “won” doesn’t mean that our ideas have permeated the entire world of computing yet, so there is still a need for a venue to discuss these kinds of topics. OSCON is more than that, though. It’s also a place where hackers (in the good sense) have a chance to showcase their work, and to exchange ideas. In that sense, part of OSCON is like a computing focused eTech. Apparently O’Reilly is no longer running eTech, which is fine – the one time that I attended, I was underwhelmed. I think that perhaps what is happening in the Emerging Languages Camp might be an example of how things might move in the future.

Of course, there’s a larger question, which is why do we have conferences at all anymore? Many conferences now produce video content of the sessions. I don’t really think there’s a lot of value in having an event to do product launches or announcements. The big thing is the hallway track, which allows for realtime interchange of ideas and opinions, and in the case of open source, provides a dosage of high bandwidth, high touch interaction that helps keep the communities running smoothly. We’re in the 21st century now. Is there something better that we can do?


1 Response to “OSCON 2010”


  • I’m not sure that I was *touting* the fact that Scala’s type system is Turing Complete so much as showing how the fact can be exploited. I tend to agree with you that when your type system has becoming a programming language unto itself, something has gone very wrong. Just look at what Pierce said regarding type system *tractability*. Your type system cannot be tractable if it isn’t even guaranteed to halt.

Leave a Reply