Tuesday Keynote
- John Gage : showed a Java powered digital camera (which is also a DHCP Web server). Nothing earth-shattering there and I still can't figure out why the fact that Java is in the device should matter to a common customer)
- Scott Mc Nealy : had his much expected and always lame top-10 list. This year was about Outlook viruses, but that's pretty much where the Microsoft bashing stopped. Litany of boring figures about why Java is storming the world, and then he invited Steve Jobs on stage
- Steve Jobs : wants to make Mac the best Java platform in the local cluster (actually, they've been trying to do that for years and repeatedly failing at it, but Steve wouldn't be Steve if he left his reality distortion field outside the room). He showed a Swing/Aqua emulation (which kind of crashed). Aqua is nice looking but Swing is as slow as it ever was, even on their "super-fast PowerPC processor"
- Patricia Sueltz : new girl on the block (wasn't there last year as far as I can tell). Probably the worst keynote I've heard in years. She's boring, non-funny, spits out never-ending logorrheas of dull numbers, sucks up to anything that has "executive" and "Sun" in their job title, and to top it all, she is extremely badly dressed. Most of the crowd started walking out at this point. She introduced partners : Bill Coleman was first, then American Express, then the Sega VP (who made some spectacularly unconvincing statements about Java as a gaming language), then Magic Johnson (who looked much more comfortable than I had expected) and then some.
Wednesday Keynote
Bill Joy
I was afraid that he might go on the neuromancien tangent that seems to have taken him lately (see his article in one of the latest Wireds) but his talk was actually okay. I did notice that he only mentioned Jini once in his talk, though, which a striking contrast over last year. If Jini's coffin needs to be nailed more assertively, that would probably be it. He did mention nano-tech, but we were spared the "Why the future doesn't need us" pathos.
Basically : while Moore's law should theoretically run out of steam in the next ten years due to physical constraints, nano-technology might give it a twenty year reprieve. An average individual (he probably means a Silicon Valley citizen) carries three devices on them, it will probably one hundred times that in ten years, which makes for interesting networking problems.
His next message was about the existence of six webs : the near Web (computer), the far Web (TV via a remote), the weird Web (voice-driven), the e-commerce Web (B2B) and the pervasive computing Web. The categories look quite arbitrary to me but, well, it's interesting nevertheless. Joy is a remarkably smart person, but he has a hard time projecting his enthusiasm halo past the end of the stage, and that's too bad.
James Gosling
His presentation skills are not getting better over time (actually, his stuttering got worse). And he definitely needed more charisma to thrill the crowd with his interest in real-time Java, despite John Gage's commendable efforts to make the talk interesting. The demonstration didn't help (two mechanical arms playing a simplistic piece of music in synch). I can't believe the non-existing demo effect of his robot escaped him. Not surprisingly, the final note on the piano was greeted with a thundering silence.
It was equally hard to get interested in the kind-of-technical explanations of what the real-time specification is about, probably because I'm not very interested in this topic, but more generally because I can't understand how someone can reasonably hope that a garbage-collected language can do better than Esterel or likes. Gosling's very words are stunning : "dealing with GC issues translates in a hideous set of constraints". Well, doh.
Even himself doesn't seem too confident in Java's ability to perform well in this domain, as his concluding remarks show : "I find the mere thought of people flying or driving Java-powered devices scary".
Gosling then invited FireDrop's CTO on stage, who proceeded to copiously pitch his zaplets. I'll write a separate rant on those later, there's a lot to be said :-)
Jon Bosak, Greg Papadopoulos, James Gosling and John Gage concluded with an unformal discussion about XML and Java. Nothing really new there, and that was a disappointing discussion given the panel.
Zaplets
In short, zaplets are "active emails". They allow you to send an email on which recipients can act and see the message kept in your inbox modified as more and more people respond.
The principle is sound, as anybody who has used an auction site or, closer to zaplets' philosophy, a service like e-vite, can testify. It's not uncommon to refresh the page in your browser just to see if more people joined and what they have to say.
zaplets want to bring this technology into your mailbox, and that's where I start disagreeing. Especially given the strong message that Sun (well, Scott) sent regarding Outlook : not everyone feels confident dealing with a mailer owning a mind of its own. Besides, checking the stored email in your inbox is not very different from refreshing your browser, so why would users change their habits ? I can't think of a really compelling reason right now, but I'm open to suggestions.
Another issue is that zaplets leave the "open" world of browsers and enter the shady kingdom of proprietary software. I can think of at least three mainstream mailers which probably gather 95% of the total industry (Outlook, Netscape, Eudora). Each of them needs to receive some enhancements to support zaplets. It's hard to tell which underlying technique they use from their Web site, but for the sake of discussion, let's assume it's plain J2EE (probably JSP + DHTML). Even this open standard is hard to obtain uniformly across the three mailers I mentioned above (you should probably not hold your breath for Outlook).
Another consideration is that Outlook already offers all these "distributed white board" features, and no matter how (un) safe these are, they are extremely widespread across the enterprise. Yes, ActiveX are used all over the place, and if you have worked a little bit with some of the COM interfaces exported by Outlook, there is no denying that it is a very powerful offer, albeit Windows-centric (which a lot of people have absolutely no problem with). The ActiveX approach is the right one, it just needs more security and thinking. Just like Scott said, "Melissa and ILOVEYOU do not need to happen". That's quite right, but it doesn't invalidate completely the ActiveX approach, and it certainly doesn't mean that the sandbox is the answer.
The bottom line is that zaplets might be the hype of the week (I'm not even sure it is actually) but they have some extremely steep technical and political curves to face, and I think they're a little late in the game to change anything in the way we use our mailers. But time will tell...