On Tue, 2007-02-13 at 14:59 -0700, Levi Pearson wrote: > I'm aware of that, though I've heard that although it's an > improvement, it still leaves some dynamic languages as second-class > citizens. I'm afraid I don't remember the details, though. A little > work on the Strongtalk VM (which was recently released under an open > source license) would produce a much faster platform for dynamic > languages than JVM would be even with JSR 292, but alas, few people > seem to be both qualified and interested in hacking on it. Strongtalk > was based on the same technology that HotSpot was, but extended > further for support of Smalltalk, which is about as dynamic as a > language can get.
>From what I've read, they are bytecodes to allow the changing of a class on the fly. Adding methods, removing methods. The problem is that this seems to sort of shoe-horn small-talk style OOP into the rigid, Java-style OOP, which is really borrowed from how C++ implemented it. Smalltalk doesn't really have a concept of a class and a class instance. Rather everything is an object that listens for and can send messages. The messages themselves are analogous to Java methods. Python, for example, merely implements a __getattr__ method on every object that can be user-defined to bring this smalltalk-like behavior to python. So I wonder how well it will solve this problem. .NET's CLR seems to have bested the JVM in the dynamic realm. IronPython runs amazingly well on it currently. We'll see how things turn out. I'd love to have a JPython that could run effectively on the JVM. Sadly right now it is lagging very far behind CPython and is much slower. Michael > > > Java 7 will be even better for dynamic langs. > > I'm looking forward to seeing how that plays out. > > --Levi > > /* > PLUG: http://plug.org, #utah on irc.freenode.net > Unsubscribe: http://plug.org/mailman/options/plug > Don't fear the penguin. > */ > /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
