Nancy has written up some "180 degrees" use cases, which she'll share more widely soon. These got Eve thinking.
We briefly discussed use cases where the requesting party needs to trust the resource owner before taking some action (potentially against resources shared), e.g. a loan officer needing to trust that the putative resource owner is who they say they are so that the (e.g.) personal attributes (resource) shared can be trusted to be associated with that resource owner "bearer". This way, the loan officer can approve a loan (which might be a second resource that the same resource owner could later share with others).
The current UMA model allows the RO and their AS to match RqP claims (not just a plain authenticated identity) against policy, and the RO can be decoupled (asynchronous) from that process. The client that the RqP uses is explicitly accounted for in the protocol, and UMA has a framework for this matching and for the RO's delegation / access granting to the RqP. But it only accounts for the client that the RO uses to interact with the RS and AS through the OAuth authorization endpoint (resulting in a PAT), and otherwise the client handling on the RO side is implicit.
The notion of the RqP needing to trust the RO and the RO needing to grant resource access to the RqP seems similar to the "decoupled" use cases, where a CSR or bank teller needs to know that Alice is really Alice before getting access to her account.
What would make sense for ensuring that the RqP could come to trust the RO "binding"? Alec will describe some work they've done along these lines in our next meeting.