On 14 September 2010 12:36, Martin Sustrik <[email protected]> wrote:
> > Major hurdle is implementing a cross platform timeout mechanism > > otherwise spin loops drive out maximum CPU time for each transport. The > > caveats of user space network protocols. > > Have you considered using already built-in 0MQ timer? It's pretty lame > as the only interval it can wait for is 0.1 sec, however, assuming we'll > limit the recovery to be 0.1 sec or higher, it can work pretty cleanly: > > If 0MQ tries to send a packet to OpenPGM, OpenPGM would report EAGAIN. > 0MQ would set up the timer and after 0.1 sec it will retry. Given that > recovery interval is >0.1s there will be no hick-ups in outgoing data > stream. > > Would that possibly work? > The timeout is more likely in the range milliseconds to microseconds for rate limiting. 0.1s can be valid for recovery. -- Steve-o
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
