[hintjens]

Yes, the over-conceptualized abstractions can be annoying. However if you read the RFCs that specify the patterns, http://rfc.zeromq.org/spec:28, http://rfc.zeromq.org/spec:29, http://rfc.zeromq.org/spec:30, and http://rfc.zeromq.org/spec:31, you will see that each pattern encapsulates more than just the routing model.

When you send() you also need to express how exceptions are handled -- what happens when there are no peers, and what happens when their buffers overflow.

Martin Sustrik did a fine job when he designed those patterns because they are (so far) watertight containers for rather tricky semantics.

[nly]

Yeah, I'm sure there are loads of technical details... but I can't help feeling there are parallels with the over zealous use of 'patterns' in OOP.

When I started out with 0MQ I felt sucked in to a world of combinatorial explosion where I always have to think about what type of socket is at the other end of of a connect()... it just doesn't feel right to me.

[angersock]

These "patterns" exist because people like you want to gloss over the "loads of technical details". :)

[nly]

Just the opposite.