Hi all,

Support for advancing draft-ietf-oauth-transaction-tokens. Review comments
from an implementer's perspective - we build agent-orchestration
infrastructure where transaction context propagation across service hops is
load-bearing for audit evidence, so this document matters to us downstream.

Three comments, none blocking:

1. Replacement semantics and audit trails (section on Txn-Token
replacement): when a token is replaced mid-transaction, the draft is clear
on how the replacement is minted but lighter on what an auditor
reconstructing the transaction later can rely on. A short paragraph stating
that the replacement chain SHOULD be reconstructible from retained tokens
alone (each replacement bearing a verifiable reference to its predecessor)
would strengthen the document for compliance-driven deployments -  in our
sector, the transaction record is frequently reviewed by a party who was
not present at execution time and holds only the artifacts.

2. The requester_context claim and downstream trust: the draft
appropriately leaves requester_context extensible. It may be worth an
explicit security consideration noting that downstream workloads MUST NOT
treat requester_context contents as verified claims about the original
requester unless those contents are independently signed - we have observed
implementers conflate "carried in a signed token" with "attested by the
issuer," which are different assurances.

3. Interaction with human-origin transactions: for transactions that
originate from a human action, the draft's model carries the subject
through sub_id. For deployments where regulation requires demonstrating
that a specific natural person stood at the origin of the chain (our case:
legal and critical-infrastructure settings under EU AI Act-adjacent
obligations), it would help to note that sub_id MAY reference an identity
assurance mechanism defined outside this document. No normative change
needed — a sentence acknowledging the composition point would prevent
divergent profiles. We have documented one such root mechanism in
draft-zagarella-verified-human-root-00
<https://datatracker.ietf.org/doc/draft-zagarella-verified-human-root/>,
offered as a composition example rather than a dependency.

Thanks to the authors -  the transaction-boundary framing is the right one,
and we intend to implement.

Roberto Antonio Zagarella
Violet Shores — VAC Protocol / Athena
https://vacprotocol.org
_______________________________________________
OAuth mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to