Re: [core] Dnsdir early review of draft-ietf-core-dns-over-coap-01

[Tim Wicinski <[e-mail]>]

Martine

On Tue, Jan 17, 2023 at 10:23 AM Martine Sophie Lenders <
[e-mail]> wrote:

> Hi Tim,
>
> many thanks your review, we highly appreciate it! And sorry for the late
> reply. The transition to the new year was very packed.
>
> On 22.12.22 18:12, Tim Wicinski via Datatracker wrote:
> > Reviewer: Tim Wicinski
> > Review result: On the Right Track
> >
> >
> >
> > I'm doing an early review of draft-ietf-core-dns-over-coap
> > for the DNS Directorate.  This is only an early review.
> >
> > Status:  On the right track, but should have another review or reviews
> > as it reaches WGLC status.
> >
> > While this draft is focused on the communication between a DNS-over-CoAP
> client
> > and DNS-over-CoAP server, the communication between the DNS-over-CoAP
> Server
> > and the "DNS Server" is done using DNS-over-DTLS, which the DNS
> community has
> > not seen implementations in use.  The whole communication between the
> > the DoC Server and the "DNS Server" should be expanded upon.
>
> In the draft, we tried to articulate that the DoC Server is just a DNS
> client (either a recursive resolver or stub resolver). The communication
> between the DoC Server and the existing DNS infrastructure (any
> recursive or authoritative DNS server) is based on common DNS protocols.
> This might be DNS-over-DTLS, but could also be classic DNS-over-UDP.
>
> Would the following figure help?
>
> +------------+        +------------+       +------------+
> |            |        |            |       | DNS Infra- |
> | DoC Client |--CoAP--| DoC Server |--DNS--| structure  |
> |            |        |            |   |   |            |
> +------------+        +------------+   |   +------------+
>                                         |
>                                         \ DNS over UDP|DTLS|...
>


Would this be used instead of Figure 1? Figure 1 with some more
explanations may be enough - let me think on this and suggest
some text.



>
> > The document talks about "DNS Server" but it needs to be more precise -
> is this
> > a DNS Recursive Resolver, Stub Resolver or DNS Authorative Server?  The
> document
> > needs to be more precise about this.  Please see
> > https://www.rfc-editor.org/rfc/rfc8499#page-16 for more details.
>
> Thanks! We will clarify in the next version.
>

Great - Stub resolvers and recursive resolvers have subtle differences
in handling resolution - especially in CPE type devices.


> > Before WGLC, from the DNS perspective a working implementation for
> observation.
> > While the authors and WG may feel that this DNS interaction will be
> > constrained to the CoAP environment, but those who operate DNS
> > infrastructure will feel this may be wishful thinking.
>
> We provide two reference implementations, one DoC client implementation
> available in the IoT operating system RIOT [1], and one DoC server
> implementation in Python [2].
>
>
AHA!  I didn't see these mentioned in the document/appendix
I have grabbed the python version and will start poking at.

In the datatracker I see y'all have links under "Additional resources"
Perhaps you can add those links under "related_implementations" ?
(can authors update these or just chairs?  as a chair I have not
actually figured that out)


I know this review is very *very* early, but once you sit down to
do document updates, please feel free to reach out.

thanks
tim

>
>