Some methods implicitly or explicitly assume exactly one account. If multiple accounts have been authorized by CAIP-25, but none "selected"/set-as-default/etc, calling said methods via CAIP-27 might produce unexpected or, worse, insecure behavior. Some mitigations that have been floated so far include:
CAIP-25 caveats/permissions/etc that would select a "default"/preference explicitly at time of provider authorization
CAIP-27 failure response with specific error message in case of ambiguity between accounts
CAIP-27 failure response with specific error message in case of zero accounts authorized for that chain
But maybe what's more needed is a use case (so far no CAIPs have a ## Use Cases section!), or a good, meaty example of a corner case where this happen? Do people know of EIPs that explicitly or implicitly define a method assuming exactly one account?
Some methods implicitly or explicitly assume exactly one account. If multiple accounts have been authorized by CAIP-25, but none "selected"/set-as-default/etc, calling said methods via CAIP-27 might produce unexpected or, worse, insecure behavior. Some mitigations that have been floated so far include:
But maybe what's more needed is a use case (so far no CAIPs have a ## Use Cases section!), or a good, meaty example of a corner case where this happen? Do people know of EIPs that explicitly or implicitly define a method assuming exactly one account?
(Moved out of CAIP-27 Discussion)