[dns-privacy] wglc feedback draft-ietf-dprive-bcp-op

[Patrick McManus <[e-mail]>]

Hi All,

Thanks for doing this work! (draft-ietf-dprive-bcp-op)

I have a few pretty small comments.

1] There is a reference to "run a .onion [RFC7686] service endpoint".

1a] The reference to RFC 7686 is not about a service endpoint - it is about
the name .onion and how clients encountering them use them. Its not going
to help with what this bcp suggests in terms of a service endpoint.

1b] As this is a BCP, I question whether this is really advice driven by
BCP. How often is this done, and when it is done how much traffic is driven
through it so that we really understand the implications of it? This feels
more like an idea than a BCP backed up by wide experience... and there is
reason to think it might fall down at scale if really adopted as a best
practice.

2] "Clients should not be required to use TLS session resumption [RFC5077]
with TLS 1.2" can be made a bit better.

2a] how does one require a client to use a resumption method (which is what
the text warns about) in any event? Maybe we should say the server should
not offer them if it does not want linkability?

2b] the explicit reference to session tickets with 1.2 leaves out a bunch
of equivalent mechanisms such as ID based resumption and 1.3 based PSKs
that create a similar issue. DoH used language like "some form of session
resumption mechanism, such as section 2.2 of RFC8446" to try and capture it
all. perhaps that's helpful here too.

3] "Automate the generation and publication of certificates" could be
changed to say "Automate the generation, publication, and renewal of
certificates" to explicitly capture the riskiest part. It should also, IMO,
include an explicit reference to ACME (as at least an example) as we've got
a lot of experience now that ACME is a good way to actively manage
certificates through automation.

4] I am also concerned that it is unreasonable to suggest TLSA in a BCP
document. What are the examples of existing DNS Privacy Services doing so
and what are the examples of the matching clients using it? At what scale?
Do we have comparable evidence about robust management of that vs the more
traditional TLS PKI and ACME and tools built for that? any experience of
tlsa with doh at scale before mentioning it as a BCP?

hth. thanks for the consideration.

-Patrick