Software design can be modeled as a sequence of transformations applied on initial specifications. In the database domain too, most engineering processes can be described as schema transformations. First, this paper presents a wide spectrum specification model intended to describe data structures at different abstraction levels and according to the current modelling paradigms. This model includes the representation of statistical data about the instances of the data structures. Then, it defines and discusses the concept of schema transformation. The properties of specification preservation is defined and applied to two important aspects, namely semantics preservation and statistics propagation. Finally, the impact of the transformational approach on CASE tools is discussed and illustrated by some aspects of DB-MAIN, a representative CASE tool based on this approach.
