Hello.

The question asked more than once in this thread has no AND or OR answer, because two states are being treated as one.

A verifier unable to check a signature.
The hop is judged on the signatures the verifier can check and the unchecked one is recorded, counting neither way. This is what lets a new algorithm be introduced without a flag day, and it needs neither OR semantics nor a requirement that every verifier implement every algorithm on the day of publication.

An invalid signature alongside a valid one, both of them checkable.
This is not an admissible outcome. The signature that verifies establishes that the content is intact, since every signature at the hop covers the same content, so the failure of the other is damage confined to that signature. A defect and an attack are indistinguishable from the message, so an exemption written for defects is available to an attacker on the same terms.

The OR argument optimises deliverability during bring-up, which is a property of the deployment stage and not of the algorithm pair. A flag leaves that choice with the signer, which is the party being checked. A sender under deliverability pressure will set it to the permissive value and leave it there.

One narrower point, on treating two signatures of the same algorithm as valid when one passes and the other returns TEMPERROR. A TEMPERROR covers a failed DNS lookup, so it is remotely inducible. Two signatures of one algorithm differ only in the key they use, so an attacker able to prevent resolution of one selector removes the stronger key from the decision while the verdict stays valid. A transient condition has an answer of its own, which is to defer the message with a temporary failure so that the sender retries. That costs an attacker a delay and gains him nothing.

Regards,
Vittorio Moccia
_______________________________________________
Ietf-dkim mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to