On 11 September 2012 01:46, Michael Haberler <[email protected]> wrote:
> Steve, sorry to come back on it - > > > Yes, a design decision for 3.x. > > it is an irregularity, so may I ask: what is gained by disabling this in > 3.x ? > It hides the situations that may cause problems. Just found a new one today on Windows: http://support.microsoft.com/kb/2639824 > > The workaround is to simply add an IPC endpoint to the socket for local > apps. You can hard hack the support back in too, it's just a socket option > on the underlying PGM socket. > > and what do I expose myself to if I 'hack it back in' besides a local > change which doesnt make it upstream ? > Multiple senders are unreliable on one host. Receivers should be ok, OS stack forgiving, and as long as congestion control is not enabled. I would think the IPC route is trivial: s = zmq_socket (ctx, ZMQ_PUB); zmq_connect (s, "pgm://eth0;239.192.0.1:7500"); zmq_connect (s, "ipc://loopback/7500"); -- Steve-o
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
