[cabf_validation] Adding Support for ACME Scoped DNS Challenges

[Wayne Thayer]

There has been an effort underway for some time in the IETF ACME working
group that resolves a significant hurdle to ACME adoption for some
Applicants. The decision has been made to implement this in a way that
requires a BR change.

Background:
It is common for Cloud Service Providers (CSPs) to ask their customers to
delegate domain control to the CSP via a CNAME record that points to a
domain name controlled by the CSP [1]. CSPs in general, and Fastly in
particular, have found that Applicants often request certificates for the
same domain name from multiple CAs. Because (unlike TXT records) only a
single CNAME record is permitted for a particular FQDN, and because RFC
8555 requires the use of "_acme-challenge" as the DNS validation prefix,
Applicants are unable to automate issuance via ACME dns-01 in this scenario.

Solution:
A new ACME challenge originally called dns-account-01 was proposed back in
2022. Last week, the fourth draft was published [2]. The scope of this
draft has expanded to include two new challenges, but I believe that the
more relevant change is that the Authorization Domain Name (ADN) is now
prefixed with TWO labels instead of one. My understanding is that this
change was made to align with the work being done to standardize domain
verification techniques [3] in the dnsop working group. Unfortunately, I
think it's reasonable to interpret BR 3.2.2.4.7 as only permitting a single
label to be prepended to an ADN: "an Authorization Domain Name that is
prefixed with a Domain Label that begins with an underscore character."

Proposal:
I would like to remove this barrier to automation as soon as possible, and
prior to RFC publication. I can see two ways to accomplish this:
1. Add the current draft spec for dns-account-01 to the BRs as a new
validation method. There is precedent for supporting draft versions of ACME
validation methods in the BRs (3.2.2.5.6 originally referenced a draft RFC)
2. Tweak the existing 3.2.2.4.7 language to allow one or two labels to be
prepended to an ADN.

I would appreciate everyone's feedback on how best to approach this and any
concerns that you may have.

Thanks,

Wayne

[1] Note that Michael Slaughter is working on a ballot that will clarify
the requirements for DNS delegation to CAs:
https://github.com/slghtr-says/servercert/pull/1
[2] https://datatracker.ietf.org/doc/draft-ietf-acme-scoped-dns-challenges/
[3]
https://datatracker.ietf.org/doc/draft-ietf-dnsop-domain-verification-techniques/
This is issue 486: https://github.com/cabforum/servercert/issues/486
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cabforum.org/pipermail/validation/attachments/20240222/7ccdeefe/attachment.html>