When we wrote that part of the OSC Spec, we were thinking of the case in which an OSC Method doesn't need to know the address through which it was invoked, i.e., "usual" cases like setting a parameter. That's why the spec doesn't mention sending either the expanded or unexpanded OSC address to a handler --- sorry about that. Why not simply always send both? That seems more general and easier to understand than a special case, at least for me.