Sunday, November 11, 2007

ooPSLA 2007 - part 3

ooPSLA 2007 in Montreal, Canada - Oct. 21 to Oct. 25, 2007
(this is a continuation from my ooPSLA 2007 review part 1 and part 2)

Noteworthy talks
Aside from the keynotes, I attended a few other noteworthy talks. In particular I want to mention three.

Martin Rinard of MIT presented an Ownard! paper entitled Living in the Comfort Zone. This is a continuation of Martin's work on what he calls failure-oblivious computing or Acceptability-Oriented Computing. In a nutshell, Martin is fundamentally questioning our notion of usable software. My interpretation of his main argument follows. Since all software will have bugs, he questions whether we can approach software construction with the explicit assumption that these bugs do and will always exist and whether we can simply make the systems more resistant to inputs that cause problems. In recent studies for various small, but heavily used systems, he experimented and collected data that seems to corroborate his hypothesis that there is a comfort zone in which most inputs to the system could be reduced to and thus have working software that is resilient to bad inputs. He experimented with both the pine UNIX mail reader as well as the commercial AbiWord word processor.

The second talk was by a truly interesting fellow named Brian Foote who, over the years, has also acquired a reputation for asking tough questions at ooPSLA keynotes and talks. Brian's presentation was very unconventional and he was introduced as the 'conscience of ooPLSA'. Having had wine and a chat with Brian two days earlier, I would have to personally agree with the preceding words.

Brian's talk is based on a paper that he wrote with Joe Yoder entitled BIG BALL of MUD. Brian introduced his talk, if I recall correctly, as "An introduction to Post-Modern Programming." It certainly is an out of the ordinary talk by a rather unique speaker. It's a collage of what appears to be a random survey of important thoughts, ideas, people, and failures in modern software engineering. However, there lies what I think is a coherent message and thesis in all of this blurry and witty exposé, and it is that software, perhaps due to its human roots, seems to have a certain inertia toward complexity and convolution. Anyone who has worked on truly successful and decently BIG software systems can attest to this tendency. Things seem to generally get worse, not better. The code becomes a big mess and market and customer pressures always seem to prevent one from ever reengineer for the better. Some techniques may help, such as Test Driven Development or TDD and Refactoring, do help but they are not perfect, they can require considerable efforts, and they can aggravate things in other ways, thus just shifting the complexity elsewhere. Maybe Brian and Martin should meet and marry their ideas. Seems like Brian has identified the problem and Martin is pointing to one, albeit resigned, solution.

The final talk I want to mention is Erich Gamma's (IBM Rational) talk on Jazz. This is the latest initiative from the people and the company that brought you Eclipse---the Rational division of IBM. Jazz is sort of an expanded version of the Eclipse platform containing all sorts of jazzy add-ons for team collaboration. For instance, Erich showed how a distributed team of developers for the Jazz itself platform is setup and how he could join one of the sub-projects. After joining, Erich could list the current TODOs, features, team members, and so on. All bug reports can be assigned to any members of the team who can also collaborate by doing instant chats and sharing white boards and other collaboration tools. Everything is transparent to members of the teams. All events and actions are recorded as RSS/Atom feeds that team members can consume in the comfort of their favorite feed reader. All in all, Jazz promises to make for distributed software development what Eclipse did for individuals... While the demo had a few glitches and the room was crowded to the max, with people seating on the floor, I left feeling good being an IBMer and hopeful that Jazz will allow distributed software teams play together like these bands I hope to see again in New Orleans.

There are still some questions about Jazz's availability and license. I will stay out of it for two reasons. First, I personally don't think that everything that the IBM Eclipse team does should automatically be given away for free, we are a for-profit company, and in this case the OSS business model seems inexistent. Finally, I am not part of the Jazz team nor do I know the details behind the project or their motivations, goals, and customers.

A lot of people raved about the 50 in 50 talk by Dick Gabriel (IBM Research) and Guy Steele Jr. (Sun Microsystems) but I did not see it. I was visiting some old high-school buddies who live in Montréal and that who I had not seen for more than 15 years. So in many ways I had my own reminisce time and was happy to get second hand account of Dick's and Guy's exploit down memory lane.

In what will surely be my final ooPSLA 2007 post (truly this time) I will complete the braindump with a recall of the LINQ tutorial, the DSM/DSL workshop, and the DLS symposium. I'll briefly mention some highlights from the different panels I attended and give a brief carricature of interesting colleagues, friends, and folks I met at ooPLSA this year. Easily my favorite ooPLSA of all previous years.

Change history (marked with strikes and emphasis)
11/25/2007: minor English updates and additions

No comments: