Cross-Network Problems with RMI

[qa41yjus]

If a previously connected consumer crashes and is restarted with the same configuration (especially with the same port number), the provider does not accept this new connection, apparently because it somehow assumes the old consumer instance still to be connected.

Instead, intended behavior should rather allow to re-initiate and resume the communication as if it was yet running like previous to the crash of the consumer.

(side note: at least in the HTTP transport the problem does not occur, restarting a consumer using the same port number does not affect the provider.)

[SamCooper]

I cannot recreate this problem. I get a two minute delay in the provider for returning a message but there is nothing I can do about that as that is TCP/IP detecting that the otherside has timed out.

In my tests, after that two minute delay, it reconnects fine and returns.