I am using remoting over TCP for a prototype distributed server application where I want to have varying multiple services exposed from each remoting server process.

In some cases I want the services running from the same process but I don't want whatever is using the service to care about that.

I am wondering is it more efficient to have multiple services in the same process going over the same remoting channel distinguished by endpoint URI/URL or should I be creating new channels on different ports for each service in the same process?

Using up ports isn't so much of a problem as the number of services will be low and the network and machine configuration is completely controlled.

Also its not clear to me if remoting sends the URI string for every single message or just at connection time, and whether if the remoting framework is intelligent enough to reduce work if calls are made on the same machine and even the same process?

Thanks in advance.