> There's an interesting discussion on > [EMAIL PROTECTED], which echoes my J2EE experience: > > http://www.mail-archive.com/general%40jakarta.apache.org/msg03376.html
Good stuff. My experiences have also been similar: simple Java stuff works very well, and the fancy distributed automagic stuff that sounds like it would be slow usually is. The obvious truth here is that if you do a lot of slow things (RPC) your system will be slow. This is why I see vendors going to lots of trouble to make sure that their "distributed object" systems try to only call objects in the same JVM. (And they brag about this as a feature!) Web systems are fairly easy to scale with dirt simple load-balancing techniques, so why waste effort on the distributed route? It's cool, but it's not good for much. Probably not a popular sentiment on this list, but if I read one more whitepaper about how someone's N-tier distributed-cache No-SQL-required point-and-click "solution" has great scalability even though it takes 5 seconds to display a simple page, I'm gonna hurl. - Perrin
