JavaOne: day 2
After some sleep, a big breakfast and loads of coffee it was time for day 2 of the JavaOne. Today's morning session was done by Oracle. Like usual Oracle did some cool scripted demo's that showed how to create a complex SOA/AJAX application by using drag & drop. Demonstrated on tools that are basically nightly builds of a technical preview, it showed once again something that may be available in like a year or so. Personally, I think pretending building an application can be done using drag & drop is a bit insulting, however as always Duncan Mills did a nice demo which still left a kind of warm fuzzy feeling. Oracle did announce some new stable products based on Java EE 5, some new developments in the appserver architecture based on Spring and OSGi, and some Tangosol stuff. Unfortunately the keynote did not did not really go beyond the buzzwords, but it's definitely something to look into.
Next we had one of the highlights of this conference: a NL-JUG meet & greet with James Gosling, aka the "father of Java". James was very relaxed and happy about the new JavaFX things coming up. The Real Time Java was also very important to him, opening exciting new possibilities to Java. We discussed a bunch of topics and closed with a group picture.
The meet with Gosling was actually planned bad so I almost missed session and a half. I did catch some Effective Java Reloaded, which basically zoomed in on some best practices using Java 5. In the afternoon I attended a talk on the new IBM JDK. This was very interesting to see how totally different the IBM SDK is. IBM did some great stuff on the class loading department and processor tuning. They finally implemented some of the new locking optimization ideas floating around devworks for a long time. Last but not least a new generational algorithm uses a hierarchy oriented spatial ordering to improve object locality (if you think that sounds like something insanely complex, that's because it is). Object locality is very important for memory performance, as IBM measured that about 45% of the CPU cycles are wasted on average waiting for memory lookups. Also IBM has a Real Time Java product offer conforming to the latest spec, which is used already in a weapons guidance systems prototype. All in all some seriously cool innovations that are ready (now, or real soon), too bad IBM does not market them well here at JavaOne.
Now the real session on JavaFX by its inventor Chris Olliver. Chris gave a nice talk using its own slide presenter tool written in JavaFX. He showed how ridiculously easy it becomes to specify UI components. I really like that JavaFX is designed with tooling in mind (for instance it is purely statically typed). For someone who knows Swing, JavaFX should be doable after a few hours (or less). Very cool, I even see somes uses for us in this in the next coming months. Make no mistake though: JavaFX still needs a huge amount of work before it is ready for use in production. I hope they can keep the promises made at the keynote.
The Intel session was a bit hard to understand. Basically Intel has joined up with Sun to make Java and Solaris a lot faster on the core duo stuff. This was actually a pretty old announcement, the only thing new was some more concrete improvement benchmarls of a rough 20% performance gain. This fits great in the JDK 6 theme for this year: Faster faster faster. After some cheesy BEA demos in how to turn 16 cores into 99% utilization using a single execution was a good thing. BEA also did some provisioning demos, which is nice, but not new. WebSphere XD does this stuff for ages, I have to look more in the details of the BEA thing to get a bearing on where this is going. With 3 parties planned, it is time to get some beer. See you tomorrow!