On 18/01/13 09:10, Pieter Hintjens wrote: > On Fri, Jan 18, 2013 at 2:00 AM, KIU Shueng Chuan<[email protected]> wrote: > >> I think there was some effort from 2.2 to 3.2 to give ZMQ a more similar >> interface to regular sockets? >> e.g. Changing the send function to return the number bytes sent on success >> instead of 0. > > This was Martin Sustrik's vision of 0MQ: "Instead of inventing new > APIs and complex wire protocols, 0MQ extends the socket API, > eliminating the learning curve and allowing a network programmer to > master it in a couple of hours. The wire protocols are simplistic, > even trivial." [http://lwn.net/Articles/369885/] > > I don't know what other people think but my own experience is that > this view of 0MQ wasn't very helpful except as an early selling point; > we suffered due to the lack of a properly designed protocol, and the > "POSIX socket" API has been clumsy to use and extend.
My goal was to allow at some point to substitute the user-space implementation (ZeroMQ) by kernel-space implementation without breaking applications too badly. Given that ZeroMQ is now evolving towards being a simple user-space async I/O library, similar to say Boost.Asio, rather than towards being part of the networking stack, using a non-standard API is perfectly OK. Martin _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
