Another nice attribute the naive implementation has is path independence. If I grab part of the globe and release it somewhere else I expect those two coordinates to determine the transform. With the non-naive implementation I can get completely different results depending on how I wiggle my mouse on the way. It's true that this gives me more flexibility, but I don't know... it just doesn't feel right to me. Google Earth does it that way too, and it has always bothered me. If I move more than a few degrees on the map I find myself either wildly gyrating my mouse to try to get back into a recognizable north-up orientation or simply clicking the compass rose to have it set things right automatically.