nn_reqrep - request/reply scalability protocol
This protocol is used to distribute the workload among multiple stateless
Please note that request/reply applications should be stateless.
It’s important to include all the information necessary to process the
request in the request itself, including information about the sender or the
originator of the request if this is necessary to respond to the request.
Sender information cannot be retrieved from the underlying socket connection
since, firstly, transports like IPC may not have a firm notion of a message
origin. Secondly, transports that have some notion may not have a reliable one
— a TCP disconnect may mean a new sender, or it may mean a temporary
loss in network connectivity.
For this reason, sender information must be included by the application if
required. Allocating 6 randomly-generated bytes in the message for the
lifetime of the connection is sufficient for most purposes. For longer-lived
applications, an UUID is more suitable.
Used to implement the client application that sends
requests and receives replies.
Used to implement the stateless worker that receives
requests and sends replies.
This option is defined on the full REQ socket. If reply
is not received in specified amount of milliseconds, the request will be
automatically resent. The type of this option is int. Default value is 60000
nn_bus(7) nn_pubsub(7) nn_pipeline(7) nn_survey(7) nn_pair(7) nanomsg(7)