Transforming UML Models to Formal Speci cations

The Uni ed Modeling Language (UML) is considered a signi cant step in the development of object-oriented (OO) notations in that it is based on some of the best OO modeling experiences. In this paper we are considering more speci cally its use in the context of high-quality modeling of complex systems. We believe that the lack of rm semantics for the modeling notations used makes, among others, the identi cation of requirements problems di cult. One approach to making UML more precise and amenable to rigorous analysis is to integrate it with suitable formal notations. In this paper we present the bene ts from a similar approach that is the result of integrating Fusion modeling techniques and the Z formal speci cation notation. The result is a technique that produces highly-structured, graphical, rigorously-analyzable models that facilitates early detection of errors. We believe that UML could gain from this experience. 1 Research motivation In the area of software speci cation we can distinguish two classes of techniques: the informal, structured techniques (ISTs) (e.g., OMT [7], Fusion [2]) which emphasize ease-of-use and understandability, often at the cost of rigor, and the formal speci cation techniques (FSTs) (e.g., Z [9]) which emphasize formality, often at the cost of ease-of-use and understandability. The lack of precise semantics for popular ISTs can lead to the following problems [4]: Understanding of models can be more apparent than real. Developers can waste considerable time resolving disputes over usage and interpretation of notation. Rigorous semantic analysis is di cult. The need for a rigorous basis for OO modeling suggests the use of traditional FSTs, where a FST consists of a notation that directly utilizes mathematical concepts, and provides mechanisms for formally analyzing models expressed in the notation. OO extensions of traditional FSTs attempt to provide a richer set of structuring mechanisms, but such extensions often result in semantics that are more complex. The integration of ISTs and FSTs provides an avenue for transforming an imprecise model into a more precise model. We propose that rather than attempting to extend traditional FSTs with OO concepts, a more workable approach would be to provide a formal basis for the OO modeling notations and concepts that have been shown to be e ective in practice. In other words, the approach we advocate is to make more precise the notations and concepts that are currently perceived as re ecting the best practical experiences in OO modeling. Formalizing such concepts and notation would deepen our understanding of OO modeling and result in OO techniques that are at least rigorous. In this position paper we give an overview of a rst attempt in this direction. Our approach for requirements modeling and analysis, called FuZed, integrates an informal OO analysis technique, Fusion [2], with a formal speci cation notation, Z [9]. In section 2 we detail our