plug test description: dealing with multiple hosts

[chrysn]

It's a minor thing but was observed in the plug test: When the client part is played by a software participant (as opposed to an embedded participant with different network stacks for node1..3), chances are that all nodes run on the same IP address and consequently need different ports.

Not sure what best course of action is; I'd probably go with suggesting dedicated ports for node1..3 to listen on, if it's not 5683 for all of them. (Plus a note on which ports are recommended for network monitoring in the tests.)

Given I'd like to run the next plug test with F-Interop support, I'll gather feedback from there as well as to whether they have good experience with any particular mode here.

[petervanderstok]

Not sure what the point is here. Do you want the RD to semantically correlate and check the registrations; for example there are two ep's with the same address and port, error 4.xx ?

[chrysn]

The point is that while things are fine in a realistic scenario (where every device has its own network stack), running several pieces of test software (a simple client, a regular client and possibly CT and lookup) on a single host means that at least two of them will compete for the default CoAP port.

The RD should not do anything about it. It's just something I'd like to make plugtest participants aware of, and possibly suggest ports to use so network sniffers can tell signal from noise during plug tests.

[jaimejim]

Not sure if this helps but if the nodes are running on the same host, virtualisation might help. AFAIK on docker you may get different ip:port per container, and you can run a CoAP Endpoint on individual containers too.