Cedric reports on Gregor Kiczales' entrance to the XML vs Java AOP syntax debate. This reminds me of the early days of C++: should we implement C++ as a preprocessor that generates C (like CFront) or should the compiler understand the language directly? Cedric's point about language independent aspects is interesting, but I think it really only makes sense in an environment like the CLR, and even then I'm not sure it will work.

James is suggesting an AOP JSR. I don't like this idea. I don't think that we understand the AOP space well enough to go standardizing it yet. And even if we did, I think it is better to allow competition among the AOP implementers. Look at what happend with XML. There was a big push to standardize XML APIs for Java -- the reasons were mostly political, and now we're stuck with SAX and DOM revisions which can only update when the JDK updates. Pull parsing? Have to start a separate JSR that may ore may not get done in time for JDK 1.5. Meanwhile people on .NET have XMLReader/Writer API's and lots of other nice stuff that can't possibly arrive before JDK 1.5 in 2004. Tossing stuff into the JCP means tying stuff to the Sun JDK release schedule, which is a bad thing. The more that I see of the .NET API's the more (grudingly) respect I have for many of them. Before I used to think that Microsoft won on tools and lost on APIs. Post .NET it seems like the reverse. Java is winning on tools (IntelliJ and Eclipse) but losing on APIs. It's much harder to fix broken APIs than bad tools. Java tools are great now, but the APIs are "standardized" and some just suck.

AOPI is YAOPP (Yet another Aspect Oriented Programming Project).