Hi Ryan, Kieran, all, Thanks for -15. I have one narrow wording question about Section 2.4, separate from the delegatee-key change and the OAuth binding sketch.
Section 2.4 presents the receipt as establishing that a particular execution was covered by the grant. Section 2.2 itself has a timing tension: a Delegation Receipt is "produced before an agent acts", while Execution Layer Evidence is "generated when a specific action is taken", and the same section classifies the receipt as Execution Layer Evidence. The verifier in Section 2.2 and the gate in Section 10 likewise operate before execution, while Section 2.3 separately discusses pairing the receipt with an action log. Consider two histories with the same valid receipt and the same successful pre-execution check (including the same signed permit decision recorded by the Section 10 gate): in one, the process stops before dispatch; in the other, the action occurs. The authorization inputs do not distinguish those histories. This does not undermine checking whether a supplied action description is permitted; it leaves the historical action facts as a separate input. Am I right that Section 2.4 intends a conditional coverage conclusion, with the occurrence and relevant action facts established separately? A possible clarification is: A validated Delegation Receipt authenticates the signed delegation terms and instruction commitment, subject to the applicable trust and key-custody assumptions. Given an action description, the relevant time, and the required policy and status inputs, a verifier can check that description against those terms. That comparison does not by itself establish dispatch, execution, completion, or an external effect. Applying the conclusion to an actual execution additionally requires evidence establishing the relevant action facts and binding them to the same occurrence and grant. The receipt alone does not supply those facts. This would retain the existing limits on free-text scope and skipped checks, and would not add per-action principal approval. Nor would it prescribe a new runtime record format or a particular observation mechanism: the action facts could come from an appropriately grounded native record or verification procedure. Would it be clearer to qualify the receipt's role within that combined evidence in Sections 2.2 and 2.4, rather than describe the pre-execution object itself as establishing the historical execution? Reference: draft-nelson-agent-delegation-receipts-15 <https://datatracker.ietf.org/doc/html/draft-nelson-agent-delegation-receipts-15> . Best, Mikhail
_______________________________________________ OAuth mailing list -- [email protected] To unsubscribe send an email to [email protected]
