[omniORB] Oneway calls and asynchronous messages...

[Charles W Collins]

Hello, I'm working on a distributed storage system project, and we want
to use CORBA for a lot of the messaging so that we don't have to worry
about marshalling objects and such.  However, in our project, we don't
need to know if a computer  got our message or not until we get a
response, and it is acceptable behavior for a few messages to get
dropped if the other machine is down.  Therefore we have tried using
oneway calls.  The problem is that when the other machine is suddenly
disconnected, and more than about 100 messages are sent, the client
machine blocks, and this isn't the way that we want it to behave.  We'd
much rather it drop the messages.

Is this a well known behavior of OmniORB's oneway calls?  Is there
anyway to tell the tcp buffer (which I assume is the problem) to discard
the messages that couldn't get through?

Thanks

Charlie Collins
Student, Carnegie Mellon University
cwc@cs.cmu.edu