A scope to be used for update could be larger than a domain of reachability, i.e. a system may allow references from a feature to another only within a part of the scope in which that reference (feature handle) has meaning, but might allow corrections to be made to anything it knows about. (Existing examples are systems which allow topological relationships only between features in the same thematic layer.)
Even at our limited current state of knowledge, we can probably suggest that our lives will be simpler if we propose that a feature handle has meaning only within one scope, and that the handle itself contains the means to uniquely identify that scope.
How we identify scopes and how we define the function which evaluates a
feature handle to produce (on demand) that feature or how we
evaluate the handle to produce some kind of scope handle, is
another matter.