invokedynamic (JSR 292)

[ben@bpfh.net]

Given all the discussion about AUTOLOAD and can, I thought some people
might be interested in this post from one of the JRuby guys about
invokedynamic, which is a proposed new instruction for the JVM in Java 7:

http://headius.blogspot.com/2007/01/invokedynamic-actually-useful.html

There's some quite interesting stuff here, about not want to totally
stuff the existing Java static type system and/or give up the security
model which prevents classes being reloaded or modified while they're
in play.

The author's general thrust seems to be that there should be two separate
sets of methods - virtual methods, which come from static typing (but should
not be called static methods, for fairly obvious reasons) and dynamic
methods.

Objects originating as compiled Java source would typically have only
virtual methods (but could have some dynamic methods via some sort of 
not-in-Kansas mechanism). Objects which started life as dynamic language
source (eg Ruby) would typically have all their methods marked as dynamic.

We now return you to your regularly-scheduled flinging session[1]

Ben
[1] Somewhat like the original end of Dr Strangelove, but with Peter 
Sellers in a monkey suit. And covered in rather more unsavoury payload.