[zeromq-dev] PUSH/PULL over inproc and nonblocking sends

Martin Sustrik sustrik at 250bpm.com
Thu Aug 4 10:46:19 CEST 2011


On 08/04/2011 10:30 AM, Ian Barber wrote:

> Is it possible that this specific case is also affected by the summing
> of the HWM? To my eyes, it looks like if either side is set to infinite,
> the HWM will always be infinite (so effectively the HWM 1 on the sender
> is being ignored):

Yes, that's right.

The behaviour mimics the TCP behaviour which has two buffers with 
separate HWMs, one on sender side, other one on the receiver side.

With inproc there's only one buffer, so the behaviour is simulated by 
summing up the sender HWM and the receiver HWM.

Martin



More information about the zeromq-dev mailing list