Simon, Raymond, All is clear now.
I was missing some initialization of the binding, and did not understand that the policy attach point was actually updated with the results of policy resolution / computation. Now I have the resolved policies on the binding, and can (if necessary) implement them at invocation time in the binding invoker. In passing - I still find this slightly surprising, as the general convention appeared to be that model artifacts simply hold the contents of the model and are not updated. Apparently that was a false impression. As for the 'if necessary' - I think Raymond's new SPIs look good, as a way of achieving maximum genericity in policy handling. Many thanks both for your help. Steve
