Carsten:

Thanks for the insight , some comments and Q..

All the coap(related) RFCs still leave open the question what, if any,
option there is to zero-touch discover (and ideally set up) discovery
in mesh networks, or for that matter any other L3 networks with lets
say two or more router hops.

> https://www.rfc-editor.org/rfc/rfc9176.html#section-4.1.1

This RDAO in RAs is an efficient way to announce an RD address that
is not on the same subnet, but unless i know of any network profile
that introduces an automated mechanism to configure this, i will
contend, that - like IMHO all other RA parameters - this needs to be
"manually configured" (which means at best you need to write some 
proprietary SDN provisioning system  to configure it).

I am also not sure i find the need to announce the RD address from
a router the most easy to deploy when the original CoAP design should
actually allow to deploy it as pure application and not router code. E.g.:
If i had a bunch of VMs in one or across multiple Cloud-DCs, i would
have some good trouble with setting up announcement and discovery via
router RDs, but no problem of using subnet "standard" CoAP discovery
via FF02::FD (and just CoAP(+security)/UDP unicast/multicast sockets
from application layer code as opposed to "where the heck is the
application layer API" RA packets.

Q: To what extent is it, or is it not possible to announce
"third-party" service information, including RD addresses just
via "normal" Coap Discovery responses ? If thats possible, then
i could much more easily figure out a distributed CoAP application
infra that just needs to have an agent on every subnet that 
would do proxy-announcements.

Especially given how we're having to build such an agent anyhow
(join-proxy).

There is some notion of proxying in the CoAP RFCs, but its rather
evasive to me. But if i for example put my DNS server with SRP
onto my registrar system (or equally a CoAP RD), then i can use
simple hop-by-hop proxying of the service-discovery for those
services (DNS or CoAP-RD).

Given how simple proxying will often give me stale information
(simple expiring mechanism), i would need to use anycast for
the service address, which is a bit of implementation pain, but
thats not network-wide, but only on those instances that would
run the services (registrar(s))...

Cheers
    Toerless

On Thu, May 05, 2022 at 08:55:18PM +0200, Carsten Bormann wrote:
> > 
> > If we had known that we'd spend 2 years in the RFC-editor Q waiting for
> > DTLS1.3, then maybe we would have done the document differently.
> > 
> >> Except that neither one works for L3 networks multicast IMHO (i am getting 
> >> no
> >> response to that request i sent meaning at least nobody knows or cares), 
> >> and COAP
> >> does not provide unicast discovery like DNS-SD from all i know.
> > 
> > RD is a core part of RFC7252:
> > https://www.rfc-editor.org/rfc/rfc7252.html#section-7
> 
> This section discusses both service discovery and resource discovery, which 
> employ Web Linking (e.g., with the Link-Format RFC 6690).
> 
> (Be careful with the abbreviation RD, as that is usually taken to mean 
> resource directory, as in RFC 9176.  RD is useful here on its own, but also 
> works with CoAP resource discovery, see
> 
> https://www.rfc-editor.org/rfc/rfc9176.html#section-5.1
> 
> Together with
> https://www.rfc-editor.org/rfc/rfc9176.html#section-4.1.1
> this is a viable alternative to relying on subnet-wide multicast to work 
> well.)
> 
> > you can do it multicast or unicast, and RFC9148 does that.
> 
> (It = CoAP Resource Discovery?)
> 
> Grüße, Carsten

-- 
---
[email protected]

_______________________________________________
Anima mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/anima

Reply via email to