> From: Steve Downey [mailto:[EMAIL PROTECTED]] > > Most objects don't work if they are made distributed without careful > consideration.
I wonder if that has to be the case. Right now, our distributed object containers are blissfully stupid. We (humans) can point at any individual class or interface and say "remote thyself!" but that's as far as it goes. The container just does what we programmers tell it to. Just like me, a hypothetical "smart container" could make some pretty educated guesses about where and when a set of existing objects should be remoted. Furthermore, it could automatically learn, over time, not only where the best boundaries are, but what are the best situations to perform the remoting/migration. Instead of deciding which classes should be remote, a smart container could decide which *objects* should be remote. And it could "learn" the answers by watching object behavior over time. Of course, architecture would still have a big effect on performance - but it does already, even in a single-machine environment. We don't hand-optimize our assembly anymore, and suspect that eventually we won't be hand-optimizing our distributed systems, either. JADE and Mobile Agents are not quite what I have in mind. The "agent" concept is very thread-centric, whereas this idea is object-centric. It sounds like EOB and AltRMI are closer to the mark. I'm going to have to take a long look, and maybe try to restart this discussion on one of their mailing lists :-) Jeff Schnitzer [EMAIL PROTECTED] -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
