What should this spec say about DPoP server-provided nonces and replay protection?

[aaronpk]

Is there anything this spec needs to say about server-provided nonces?

• authorization server provided nonce
• resource server provided nonce
• replay considerations

My suspicion is that there isn't anything unique to this spec that would dictate whether these are required, and that they are still going to be up to implementations to decide whether are necessary. But it at least wouldn't hurt to mention these mechanisms for a reader of this spec who isn't familiar with the details of DPoP.

[mcguinness]

Agree with your instinct. DPoP nonces and proof-replay are RFC 9449 mechanics that this spec doesn't need to mandate or constrain.

There are two things I think might be worth saying explicitly for a reader who isn't deep in DPoP

1. The AS-provided-nonce mechanism (RFC 9449 §8) applies at two endpoints in this flow: the IdP's Token Exchange endpoint and the RAS's JWT-bearer token endpoint. Each maintains its own nonce state; nothing crosses between them.
2. DPoP proof replay protection (RFC 9449 §11.1) is per-recipient. htm/htu ensure each proof is scoped to a single endpoint, so the IdP and RAS track DPoP jti independently.