Jeff, Jijie, all,

Jeff, this is a useful distinction, and I think it exposes a real role
collision in the current text. I read RFC 7523 §3 a little differently
on one point: it allows `sub` in an authorization grant to identify
either the resource owner or an authorized delegate. An agent can
occupy that slot, but only if the profile says clearly which mode is
in use.

There is also a concrete mismatch we can fix now. AIC-JWT §5.2 defines
the DA payload with `agent_id`, `principal`, `requested_lifetime`,
`ts`, and `nonce`, but without RFC 7523's required `iss`, `sub`,
`aud`, and `exp`. Section 10.2 then presents that DA as an RFC 7523
`jwt-bearer` grant, so those sections do not interoperate yet.

I would fix the assertion first, then define the issued token by mode.
In `representative` mode, the resource owner or principal is `sub`,
and the agent is `act` or `client_id`. In `authorized` mode, the agent
may be `sub`, with the signing authorizer carried separately. If the
accountable operator is neither party, it needs a third binding.

Omri's current token-exchange draft already separates the exchange
subject from the requesting party and may be a useful model.

Best,
Dr. I.,

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

Reply via email to