[zeromq-dev] Is there already native "high level" connection pool?

[Stefan de Konink]

On 11/27/12 14:57, Ian Barber wrote:
> So, for an example you might want to look at how Mongrel2 handles this
> kind of thing. Messages flow across ZeroMQ with a connection ID and a
> body, and get written to the appropriate HTTP connection. So, basically,
> don't use SND_MORE to handle the stream, send each part individually,
> and include a connection identifier as a message part within each one.
> As Michel says, multipart messages are delivered atomically so they wont
> give you the behaviour you are looking for, pooling or not. 

So the messages for the solution with the lowest latency look something
like:

[Envelope + MessageId, DataPart]

Where the triplet is send atomically and the workers below it
reconstruct the messages as Michel suggested? (Given that the messages
arrive in incremental order.)


Stefan