I figured since everyone else was sending 'me' messages, I may as well
also :)
I am in Colorado Springs, happy, and moving to a new (local) job in two
weeks or so. I'm working on Classpath, but kinda slowly, since I'm trying
to get a response from Chris on this gnu.vm proposal and on how we can work
with them and get a Japhar+Classpath version rolled out.
Finally, I need to get back to java.beans in a bit. The first run at
Introspector works like Sun's, but the extra design req's I threw in there
to make the implementation more predictable are slowing the implementation
down to something like twice as slow as Sun's on the Sun JVM. While greater
predictability is desirable, I don't believe it warrants this kind of
slowdown (mine requires basically two passes, though by optimizing in
various ways I could probably bring the speed differential down to 1.5 or
so). Most people are used to the unpredictability of the Introspection spec
(may not even realize it's there) and design their classes so that
ambiguities don't arise. I will assume they do that and say "it's your own
fault" if they design ambiguities into their classes. I will, of course,
document this fact in the javadocs.
--John Keiser
