This is one of the reasons why i would prefer to have a NodeChangeEvent object.I second jms for this, could you also put 'protection by sending cloud id/name' to each message.
Yes i understand this, maybe it has to do with the application server.This is a danger point of the current implementation. Also Andre's problems at the start of the thread are not (imho) related to multicast itself but a until not defined/found problem of why a burst and more important why is this a problem on the clientside (since they should be really quick).
But, there are better ways to handle asynchronous requests instead of starting a thread for every request you receive. You could think about a producer / consumer mechanism with a producer that puts the incomming requests in a queue.
Rob
