On 04/21/2016 12:45 PM, Hugo Landau wrote:
> https://datatracker.ietf.org/doc/draft-landau-acme-caa/

This mostly looks good to me. A few comments:

A little while back when we first discussed this, I proposed using ACME
account URLs rather than thumbprints. At the time I could have gone
either way, but thinking about it more, that seems like the right level
of abstraction. The point of the record is to authorize a specific
account, rather than a key. This also means that the CAA account binding
spec doesn't have to be so tightly coupled with the details of how ACME
request authentication works.

Also, I think the security section should mention that the CAA model of
evaluating from the leftmost label means that this is not an effective
means of controlling issuance in organizations where DNS for subdomains
is delegated to parties that may configure a different CAA record.

The security section could stand to be a little more pointed about the
risks for non-DNSSEC-enabled domains: Network attackers that could spoof
an ACME validation response could also spoof a CAA response authorizing
an attacker-chosen account key.

Thanks,
Jacob

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

Reply via email to