It seems to me that one possible base for modeling visualization transforms would be a camera model. In such a model the meta data must be stored would consist of changes to camera position, orientation, and field of view. Recording this would require development of an xml scheme for specify both value in each of theses, and a way to represent changes to an existing value. For position and orientation the same representation would probably work for both a value and change in value. The schema for field of views would be more complex. It is also work noting that this would still be a subset of possible transforms. Another type of transform is the determination of how data is rendered. 

Posted by: Donald Johnson