Hi again,

On 15 Nov 2026, at 14:21, Philip Homburg <[email protected]> wrote:

> As far as I know, the following two statements are true:
> - with draft-ietf-dnsop-multi-alg-rules it is allowed to have a zone
>  where some signers sign the zone with ECDSAP256SHA256 and other
>  signers sign the zone with ECDSAP256SHA256 plus ML-DSA-44.
> 
> - with draft-jabley-dnsop-local-signing-algorithm-policy, when a validator
>  implements the policy described in the draft and receives an RRset signed
>  with just ECDSAP256SHA256, it concludes DNSSEC bogus.
> 
> That leads me to conclude that these two draft are in conflict: large
> scale implementation of one of these draft would make the other
> draft impossible because it would lead to DNSSEC bogus results.

Oh, I see what you mean, now. A response that only includes signatures from the 
P-256 signer would be identified as an attack by a validator that wants to 
protect against ML-DSA-44 signature stripping. Your ZSA idea also wouldn't help 
in this scenario: if an ML-DSA DNSKEY has ZSA=1 and one signer only signs with 
P-256 then the RRSIGs generated by that signer only will be insufficient for a 
validator that includes the ZSA logic.

> But maybe I'm wrong and I misunderstood the drafts.
> 
> I support the goals of both drafs. It is just that as far as I can tell,
> as the drafts are written today, they cannot co-exist.

Presumably they can co-exist if zone administrators choose their signing 
algorithms appropriately and avoid the case where one signer doesn't sign with 
a PQ-safe algorithm while others do. Maybe this is just a more general PQ 
consideration (as the sole case we care about today that has downgrade 
concerns) and not something that multi-alg-rules needs to care about.


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

Reply via email to