To acquire the complete description of an arbitrary object's surface using a range camera, multiple images of the object from different viewpoints must be combined. This paper presents a solution to the problem of determining how to position the range camera so as to sample more of the object's surface while making sure the new image can be registered and integrated into the (incomplete) model of the object. Requirements for the solution to this problem are 1) all portions of the surface which can be scanned will be scanned; 2) the solution will generate a set of views which when combined will converge to the object's surface in a minimum number of images; 3) all surfaces of the object should be scanned with at least a given minimum confidence; 4) the solution will be robust in the sense that if it chooses a viewing position for the camera the resulting image will be successfully registered and integrated with the (incomplete) model or it will announce that now such view exists. We present a framework for such a solution and argue that it satisfies conditions 1 and 2. This sets the foundation to extend the present solution to satisfy conditions 3 and 4.