From a web app ( and from any client code ) , it s simpler if we can just instantiate a client anytime, in any context, to connect to a server ( for which we have a known configuration, ie hostname:port or some similar URL, defined in a config file ).
Currently we have one client setup when using the context manager, that we need to copy around and keep in context. This came to be because of our old pipe that we used as a communication channel.
Now that we are using our internal ZMP ( ZMQ comm channels ) we can improve the design to allow users to create client anytime, provided they know how to access the server process(node name - or maybe we should directly expose and use ZMQ URLs).
From a web app ( and from any client code ) , it s simpler if we can just instantiate a client anytime, in any context, to connect to a server ( for which we have a known configuration, ie hostname:port or some similar URL, defined in a config file ).
Currently we have one client setup when using the context manager, that we need to copy around and keep in context. This came to be because of our old pipe that we used as a communication channel.
Now that we are using our internal ZMP ( ZMQ comm channels ) we can improve the design to allow users to create client anytime, provided they know how to access the server process(node name - or maybe we should directly expose and use ZMQ URLs).