Here are some notes related to this year's JavaOne conference. These notes and the JSR-121 Birds of a Feather session (BOF-2436) slides can be found here. I still haven't found usable archives from the conference and will have to add those as URLs to the page above when they become available and I'll send a note to the interest list at the same time.

The day before the BOF I got registered and my wife and child were taken off by a relative to see the sights north of the city. I visited the java.net booth and got some details about the possibility of a Sun sponsored JSR using their infrastructure (Java Server Faces project, aka JSR 127, is already blazing this trail!) and bumped elbows with others (hopefully the Sun booths will be more widely spaced next year!). Around this time I noticed somebody had loaded JDistro into an otherwise empty booth and that made me giggle.

The big ("duh") revelation this year was that having a BOF earlier than the final time slot for a night (as we had last year) resulted in a drastically different dynamic. The hours of post-BOF discussion of last year were compressed into just a few minutes as we were chased out of the room by the next BOF session and then most folks had some place to go anyway, causing hallway discussion to be very brief. If anybody feels short-changed they should feel free to send mail to the interest list or direct to me at the address below.

There were a few less people at the BOF this year and it was in a smaller room, although somebody today mentioned conference attendance was higher than last year. I didn't exactly pump up expectations leading up to the conference and I think the "Sharing the Virtual Machine" technical session (TS-3207) earlier in the day might have been a substitute for some of the folks that would otherwise have attended the BOF. Likewise Bernd Mathiske had a full technical session (TS-1634) partially overlapping his material and some conference attendees no doubt realized that.

In the VM sharing session Greg Czajkowski discussed experiments exploring JVM metadata sharing and resource management related to the multitasking virtual machine ("MVM") research platform at Sun. Most of this would be familiar to folks that have read the JSR-121 bibliography materials but there was some new material, in particular experiments involving multiple appserver instances within a J2EE environment. Some of the latter work (done with academic collaboration) has just been published as Sun Technical Report TR-2004-135. This reminds me I need to get back to updating the JSR-121 interest list bibliography.

For the JSR-121 BOF, leading up to the conference I thought of some outrageous demos and by degrees realized they just wouldn't play for various reasons and as it turned out the 50 minute time constraint wouldn't have allowed for this anyway. But a simple demo of some some sort seems like a given from this point on and if Nextel's wireless broadband is available in San Francisco we might have some real possibilities with just a laptop.

My presentation was mostly familiar overview for context the new parts are really just bullets made out of material shared with the interest list in past months. Oh, my first item was to explain why Doug Lea wasn't present, which was simply that his teaching work this summear didn't allow for him to make it to the conference this time. The slides I presented are here. I should repeat a general remark I made at the session, which is that part of the "Next Steps" depend on my management, the JSR-121 expert group and various other factors, so this is more or less a wish list on my part (but a list that seems to be in tune with others).

Bernd Mathiske had flown back from Germany the night before and was masterful at disguising jet lag. For the session he shared some details of adding VM multitasking to Sun's Monty JVM to enable application switching and concurrent execution and he demonstrated this on a Sharp SL5500 PDA (note that this enhancement of Monty does not expose any APIs, to enable a proper rendezvous with JSR-121 in the future, probably via a new JSR-185 roadmap).

As expected, Bernd was also very candid with us about open issues with respect to implementing JSR-121 with CLDC. One detail that sticks in my memory is that the size of the initial implementation of isolation within the Monty JVM was a major fraction of the size of the rest of CLDC! Bernd went on to explain how the implementation got optimized and shrunk properly, but this gave the audience a real sense of the API design struggle we face. Bernd's slides are here. After the session Bernd shared his observation of expressions of consent by BOF attendees when he lobbied for having fewer and simpler APIs. One attendee also voiced a desire for simplicity to me after the session.

Pat Tullmann pointed out one observation that came with the Q&A which is that even without resource management APIs, per se, the isolation APIs provide some degree of resource isolation and a bit of control by virtue of that (i.e. at a minimum forced reclamation is possible).

Speaking of Pat (JSR-121 EG member representing U. of Utah) I should mention that Matthew Webster of IBM (the other JSR-121 EG member at the conference) sent his regrets as he had a conflict and couldn't make the BOF.

The day after the BOF I had nice conversations about the possibilities of using isolates as containers within a telco OSS (Operations Support Systems) environment and there will probably be more to report about this in the future. Then I had to grab my luggage from the hotel and catch a bus north to rejoin my family.

Best wishes,