Wish You Were Here!

Some hot news from the OSGi world congress. I am writing this while I should pay attention to the presentation, but how else can I keep you posted?

It is amazing how many people are using OSGi in a surprising number of application areas. Additionally, it looks like it is hot in the academic circles. But the best part of the congress is to speak to so many people that radiate so much energy about someting I have cared about for so long.

Yesterday we had the standard R4 presentation from BJ Hargrave and me. A blood nose made it rather more exciting than your run of the mill presentation, OSGi presentations rarely end with a blood bath! Of course we ran out of time before we had discussed all the nice additions of R4.

In the afternoon we had planned "The trial of CPEG" panel. This was a first for OSGi: humor! Adapted from the infamous trial of the gang of four at the OOPSLA. The trial was a bit scary for all the actors. Especially for Dave Marples and Richard Hall because they had to wear wigs! It was painful for me sometimes to get confronted with all those indictments; Richard drove them down hard. Many times I was only allowed to say yes and no. "Is Java more efficient than C? Only answer yes or no". Hard! Especially the witness that kept harassing me about XML and JAR files in an incomprehensible way. The judge, Dave Marples, had shown a lot of bias towards the prosecution but allowed us to be acquitted by the jury in the end; only to turn around and commit us to 6 months of writing Visual Basic 6 ...

At the end of the first day I embarrassed Jeff McAffer by saying that he was going to demo OSGi on the NSLU2 (the slug). He had told me he had J9 and Eclipse Runtime running on the slug. Sorry Jeff. He took sweet revenge today, together with Thomas Watson, by giving a fantastic demo (though he chickened out by using a Flash demo that had made the night before !) showing off Eclipse on the slug. They used a flash USB stick inserted in the slug as his Z: drive (it runs Samba). He copied Eclipse and J9 on there and then telnetted in the slug and ran the Eclipse runtime. Then they created an Eclipse bundle project on Z:, used the Hello World example and ran it in less then 3 minutes! Anybody that has ever done embedded development must realize how far we have come since burning Eproms and toggling LEDs to find out where the bugs are ... Watch this space because we will use the slug for lots of demonstrations in the future.

A couple of weeks ago I explained one of the coolest features of Release 4, the security model with the signing and bundle permission files. From that article, it should be clear that it was the first time I saw how the fine grained security of Java was feasible. Still, creating the minimum set of permissions that a bundle uses is a non-trivial task: tools are needed. To my pleasant surprise, Ted Habeck and Marco Pistoia from IBM demoed the "Security Workbench Development Environment for Java", that did exactly that. The Eclipse plugin can statically analyze class files and JARs and create the minimum set of permissions. The plugin was beautifully integrated with Eclipse, linking any permissions directly to the source code. With a few additions, this could be an invaluable tool for operators.

Well, there was so much more happening that the blog would become too much. All of a surprisingly high quality. My head is still spinning of all the exciting ideas that are popping up here. I feel sorry for the people in the audience that just came to sniff at OSGi, some of the stuff is really hard core OSGi. Next year we definitely need an introduction workshop as well.

So, I have to start paying attention to the business part of the congress. Sorry you were not here!

    Peter Kriens

posted by Peter @ Wednesday, October 12, 2005