ph at imatix.com
Fri Sep 14 08:46:11 CEST 2012
On Thu, Sep 13, 2012 at 7:57 AM, MinRK <benjaminrk at gmail.com> wrote:
> I don't think it's too much work, since the first implementation did exactly
> this, if I read correctly. There was an issue with the implementation being
> made via setsockopt, which prompted a cleaner interface by adding a
> dedicated method.
This isn't quite how it was. The thread in question has the title
"Strict aliasing problem".
The original API was just a single callback, not attached to anything.
We attached it to ctx to make it more coherent with the API that was
emerging at the time. There was never any suggestion to attach it to
sockets, though the idea clearly makes sense.
I'd suggest, since this API has not been released in a stable release,
patch it now to make it work on the socket level, and (importantly),
try it in some real use cases to see how it flies. The fact we're
discussing this after several months hints that perhaps no-one really
needed this function in the first place. (In which case I'd remove
More information about the zeromq-dev