On Sat, Jul 24, 2010 at 4:38 PM, Martin Sustrik <[email protected]> wrote:
> My rationale for hiding the info is that the semantics are not > consistent across different topologies... Understood, but this is not blocking discussion in areas such as subscription upstreaming, which make sense only for a TCP unicast topology. Looks like a perfect abstraction of all transport mechanisms is not possible, nor needed. I'd suggest that the architect knows when it's worth doing transport-specific stuff. > Can be done. Of course, it means splitting the 0MQ codebase into two > parts, meaning a *lot* of both design and implementation work. So you could build all patterns on top of raw sockets, and allow user space patterns as well? This might be worth considering for 0MQ/3.0. Anyhow, I think this is a better answer than "just use TCP sockets". -Pieter _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
