In method B3DSceneMorph> > drawAcceleratedOn:  we calculate the bounds of the 
morph by transforming them into global coordinates and intersecting them with 
the screen extent. The resulting rectangle is handed to the renderer via 
method #bufferRect:.

Later in method B3DSceneMroph> >renderOn: we use the method #viewport: to tell 
the renderer its viewport bounds. But this time no transformation or 
intersection is aplied to the morphs bounds.

So if there is anything about the bounds handling in the first case it would 
later on be un-done, wouldn't it?

Martin