On Tue, May 03, 2022 at 11:50:10AM -0700, Stuart Cheshire wrote: > > I guess this could happen when we come up with different > > protocols all running on top of coap (given how the service instance > > port number is only bound to the coap layer). > > That would be a bad design. > > The service type says what a service does, not what engineering decisions > were made in its design.
I hope we are in violent agreement (see below). > Advertising a “CoAP” service makes as much sense as advertising a service > type called “ascii” or “utf-8” or “xml” just because the protocol happens to > use those things. > > See <https://datatracker.ietf.org/doc/html/rfc6763#section-7> > > Particularly read the part that says, “Wise selection of a Service Name is > important...” The service name would not be CoAP, but e.g.: brski.rjp (service point for boostrapping of key infras). The TXT proto= key would need to be a list of 1 or more protocol variations, if those protocol variations happen to be able to (or need to) operate across the same UDP or TCP port - which is something that could happen for protocols operating on top of COAP or HTTP. For example proto=est-coap,cmp-coap, if we have two coap protocols, such as one based on EST (rfc7030), and one based on CMP, both able to operate across the same COAP (UDP) port. Service is the same, just encoding of transactions and data-structures differs. Would that be appropriate ? Cheers Toerless > Stuart Cheshire -- --- [email protected] _______________________________________________ Anima mailing list [email protected] https://www.ietf.org/mailman/listinfo/anima
