Should see if the zeromq's proc is a worthy queue implementation

[fquinner]

See http://api.zeromq.org/4-0:zmq-inproc for details. Will need performance tested against existing wombatQueue to see if it's an improvement.

[fquinner]

OK if you look in https://github.com/fquinner/OpenMAMA-zmq/blob/master/src/queuebench.c, i have thrown together a test framework for this.

I couldn't really get any numbers for fan-in because in my test, multiple producers proved difficult as you couldn't bind to the same uri more than once... Unfortunately this means this implementation will require locking in the application bridge itself because zmq sockets are not thread safe.

So if you have to lock on both sides anyway, the inproc queue is in fact slower. I can see how it would be much faster for a specific use case where you know you have (say) 2 threads which need to receive messages and that's it, but for a mamaQueue implementation, you have to write for the general case, and when you do that, wombatQueue actually works out faster so we'll be sticking with that for the bridge.