On 11/06/2026 17:02, John Levine wrote:
As I said, the intention was for resolvers to recover NXDOMAIN. If
they don't do that, I'd say it's a bug, and I'd rather fix the
resolvers to maintain the behavior they've had for the past 40 years
than try to add workrounds in every application in the world that uses
the DNS.

I agree on the intention, but the language in RFC 9824 is unfortunately quite weak to allow reporting this as a bug. I'm skeptical that reporting this to every DNS resolver that exists out there, including proprietary ones, citing text that says that a resolver "could" modify the response would work.

Plus, this would leave two cases unsolved. One is the case of "DO flag without CO flag". As I understand it, you can't respond with NXDOMAIN in that situation, as that would imply standard non-compact NSEC/NSEC3, and it's therefore up to queriers to read the NXNAME bit. Alternatively, all DNSSEC-enabled clients would need to always send also the CO flag, to signal they understand Compact Denial. So the next step would be filing a "bug" against every DNS library that exists. And once you achieve that, you would still have no guarantee it would work, since the CO flag is optional.

Did I get this right? If so, I still think that RFC 9989 is at least
missing some guidance for developers:
If we are going down that road, you need to file an erratum for every
RFC that mentions NXDOMAIN.  But please don't.

I personally see no harm in new RFCs relying on NXDOMAIN that come out after RFC 9824 to have a sentence that says that implementations should consider RFC 9824 for reliable detection of non-existent domains. As a developer, I would certainly find it helpful, especially with how significant nameservers adopting RFC 9824 are.

For anyone that may be interested in how to reliably detect non-existent domains, I'm sharing a small Python code snippet that checks for NXDOMAIN, uses DO+CO flags, and extracts the NXNAME bit. Unfortunately, this deviates from the definition of non-existent domain in RFC 9989.

https://gist.github.com/matteocontrini/8693c69fde4e9e349fd3023f61230ab3

--
Matteo


_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to