A few years ago I devised and implemented this zeromq scheme<https://docs.google.com/drawings/d/19fkuumKbSJItjAZ6RrsOYpSmA0lOZEAN8Lc2naaI-R4/edit?usp=sharing> to facilitate cloning and transmission of past-accumulated and live-updating time-series data from multiple sources to multiple selective consumers. At the time the sources were running on Windows, the consumers on Linux or Windows. I was unable to build CZMQ on Windows so it was necessary to re-invent the wheel and code up my own socket handling class and multipart-message class in C++. This scheme is in use & works reliably day-to-day.
Times have now changed; I'm able to build & use CZMQ on Windows & Linux. I would like to revisit the scheme, update it to SERVER-CLIENT socket types (ROUTER-DEALER is to be superseded) and implement it all in CZMQ, also adding authentication and encryption ala IronHouse<https://github.com/zeromq/czmq/blob/master/examples/security/ironhouse.c>. My questions are: 1. To implement this scheme the most demanding technical challenge is the accumulating, preserving & switching the prepend order of routing 'identity' frames such that the ROUTER (or SERVER?) can direct the message to its destination. Can this be accomplished with CZMQ; can multiple accumulated identity frames be accessed and prepend order manipulated? 2. SERVER-CLIENT sockets at present don't support multipart messages so zmsg_encode () & zmsg_decode ()would be necessary; will this interfere with access to routing identity frames? 3. Is the scheme/pattern of value/interest to the zeromq community, would it be worthwhile to collaboratively redevelop it on Github? Stephen.
_______________________________________________ zeromq-dev mailing list [email protected] https://lists.zeromq.org/mailman/listinfo/zeromq-dev
