Today's session began with a keynote from Jeannette Wing. It was a kind of pep talk to encourage collaboration with other fields, as well as high-risk, high-yield research topics. It was moderately interesting to hear things about NSF, since she's currently the assistant director of the CISE directorate, but otherwise, I'm not entirely affected by speeches that are calls to arms and whatnot. The presentation itself wasn't all that bad, though someone commented to me that the higher level position you are, the more vague your answers get. Totally amusing, yet totally true.
Following the keynote was the morning research track, which focused on reliability and monitoring. None of the papers really stood out to me, so I won't go into it, but you can see the research track papers here.
After lunch was another invited talk by Gerard Holzmann, who discussed the use of formal methods in software development (particularly in spacecraft, where correct software is obviously vital). For some reason, all of the larger rooms have their lights dimmed, so I was dozing off, but the basic idea was that after the initial setup, formal verification software is easy to use and helps a lot, but people are scared to use it for some reason, when they shouldn't be. When asked about the tradeoff between testing and formal methods for non-safety-critical software (such as GNOME desktop), I don't think he gave a very straight answer. He of course knows that the answer defaults to testing, but as to whether or not formal methods are truly beneficial and worth the time to understand, that was left unanswered.
The research track in the afternoon was on software tools and libraries. The sound of that to me just doesn't sound particularly researchy, but I was one of the volunteers assigned to it, so it didn't really matter what I thought.
The first talk was on IMP, a project from IBM to greatly simplify IDE development. Extending Eclipse is supposed to be comparatively complex, which is why the project was born. Essentially, you must implement a lexer/parser at minimum, and then add IDE services in at will. I know little about either project, but it sounds somewhat like Visual Studio Shell, which was released with Visual Studio 2008. It was an interesting talk, but not really from a research perspective.
The second talk was about bridging the gap between Java and C debuggers in order to debug JNI code, which essentially provides a unified stack in order to get a full stack trace when debugging problems that occur in JNI code. The content was mildly interesting, but I don't exactly deal with JNI, so it wasn't all that applicable to me. On top of that, the presenter was a little overbearing, and I'm certainly not the only one who thought the same.
The last presentation was presenting C#'s task parallel library. Even if my research is mostly in Java and I otherwise mostly try to use Python, I still consider C# to be a good language (and it's unfortunate that most researchers opt for the JVM). I never really read into the parallelism that is shipping with C# 4.0, so it was interesting, but again, it didn't feel very research-like to me. Seeing Parallel.For and futures in C# is nice, but I can't imagine it being very complex, conceptually. Rather, it was somewhat like Sunday's tutorial, where they were implementing already-known techniques in a pure OOP language.
This evening was the big, Hawaiian OOPSLA dinner. Unfortunately, I think Disney got their ethnicities mixed, because about half of the food was Chinese and the other half was American. Oops. For the curious: pork dumplings, crab rangoon, fried rice, vegetable stir fry, pulled pork, lemon chicken. Yeah, what part of that is from Hawaii? Apparently there's dancing at every OOPSLA, which was a pretty amusing thing to see. Other than that, the evening wasn't all that exciting. Since it was difficult to find a table, all the Purdue people ended up at the same tiny table with Tyler, from Iowa State, whom I met at the summer school. Not being a social butterfly and all, I feel awkward just walking up to strangers to talk to them, so I guess I missed out on a good opportunity. In fact, I haven't really met many new people at all at OOPSLA, just talked to some Purdue people more than I had at campus and meeting several of the summer school attendees that made it to OOPSLA.
Takeaways: from the research program, not much. I've had more interesting days. The formal methods talk leaves questions opened to be answered, but I have no real desire to enter that field, so I guess someone else will have to answer them. Though things like model checking are fairly interesting, I am certainly not one who indulges in formal methods and verification.
Tomorrow is the last day. It's been a tiring, but overall fun experience.
No comments:
Post a Comment