Carsten Bormann <cabo@xxxxxxx> wrote:
> Right, versions are massively overrated. (Many people think they can
> transfer the knowledge they have gained about software versioning to
> protocol versions. Totally wrong.)
I agree.
Further, we don't do enough advance work in "version 1" to say how we will
use minor protocol numbers. They are almost always useless in my opinion,
but it's taken a long time to get to that view. Minor extensions must always
be accomodated without protocol version changes.
> The reason I’m replying is that, for EST-CoAPS, the situation is even
> more complicated, because CoAP has a standard discovery mechanism
> (which is incidentally fired by /.well-known/core), and EST-CoAPS also
> defines everything that is needed to make that work for them. So there
> already is an established way to obtain the server-defined entry point
> URIs to the various services. The /.well-known/est stuff mainly seems
> to be there because there is a belief that implementers in this space
> want to nail down everything and won’t do the discovery.
Yes, I've pushed to not sit on the fence, here, and lost :-)
Either we always do discovery, in which case, we have no need for
/.well-known/est, or we can operate on never doing discovery, in which case,
we don't need discovery. It's clearly a code space vs network round trips
argument.
I think that once we have a standardized CBOR reply for CoAP discovery, so no
parsing of ASCII strings, that discovery will become more popular.
WOuld you say that in a CoAP space that protocol enhancements are simply
indicated by the discovery returning new end points. We can name
them /2/blah if we want, but /rfc1234/blah would be equally valid/useful.
--
Michael Richardson <mcr+IETF@xxxxxxxxxxxx>, Sandelman Software Works
-= IPv6 IoT consulting =-
Attachment:
signature.asc
Description: PGP signature
