Re: [core] CoRE@IETF100: Summary

[Carsten Bormann <[e-mail]>]

On Nov 20, 2017, at 13:55, Hannes Tschofenig <[e-mail]> wrote:
> 
> Hi Carsten, Hi Jaime,
> 
> On 11/17/2017 06:45 AM, Carsten Bormann wrote:
>> ## In IESG processing
>> 
>> * draft-ietf-core-coap-tcp-tls is in IESG processing, waiting for
>>  Mirja's DISCUSS to clear [it is now clear that this will be in
>>  December after her vacation]. 
> 
> It is unfortunate that she went on vacation without clearing her DISCUSS
> during the IETF week,

It is, but then ADs are not superhuman and need vacations, too.

> which means that we will not be able to finalize
> the document before xmas as anticipated during the CORE WG session @
> IETF#100.

I’m not sure how that follows, although indeed it will require a bit more attention to achieve this.

The formal steps are:
— Mirja clears her DISCUSS;
— The responsible AD (Alexey) verifies that we handled all the COMMENTs reasonably;
— The document is approved;
— We will ask Alexey to then request expedited processing from the RFC editor.

Step 2 can be parallelized with Step 1.
If all stars align we could have an approved document on, say, Klaasohm, er, Itsenäisyyspäivä (Dec 6).

> The first interop on Saturday had 2.5
>>  implementations (libcoap, augustcellars, and coap.me), with a CoAP
>>  GET performed on /.well-known/core (which exercises a lot of the
>>  machinery already).  Editorial input from the interop should lead to
>>  a -11 soon.
> 
> Is there a write-up about the interop testing available? What exactly
> was tested?

Mostly framing, connection setup with initial CSM exchange, and a simple GET.

> Is the code available as open source, maybe in a dedicated
> branch?

The code for libcoap is already mainline (branch “develop”); please see https://github.com/obgm/libcoap/pull/113
The code for both augustcellar and coap.me is evolving and AFAIK not yet in a public repo.

> Could we plan to do another online test in December?

Absolutely!

We need to set a date for that — please indicate when it would be convenient for you.
We also should work on testcases (the only one so far was a GET on /.well-known/core).
See https://github.com/cabo/td-coap4 for an example of how to do this.
(We probably don’t need as many test cases as for UDP CoAP, but could simply port some of them over and should at some point have at least one for each feature of CoAP-TCP.)

> I guess the 2.5 implementations refer to two client & server
> implementations plus one client or server implementation? Is that correct?

Both libcoap (C) and augustcellar (.NET) are client and server.
coap.me (Ruby) isn’t quite there yet; we only tested the client code in Singapore.

Grüße, Carsten