Conceptual classes and system classes in object databases (Q1386410)

Semantic data models, proposed at the end of the seventies, introduced a new quality in the design of database applications. These models offer powerful data abstraction mechanisms providing: (i) a higher level description of the database objects, closer to the real world objects; (ii) the required flexibility in reflecting the changes occurring in the real life. However, they mostly resulted in database design tools rather than in complete database management systems. As systems, they remained at the stage of research prototypes, since semantic data models were never implemented efficiently. In the late eighties, a new breed of databases systems, based on the objectoriented paradigm, was proposed to overcome the limitations of traditional database systems in new application areas, such as CAD/CAM, artificial intelligence, office systems and so on. Objectoriented database systems, such as GemStone, Orion, Iris, O2, Starbust, ObjectStore arose the interest of the industrial community for their level of engineering and performance. However, although object database systems are supported by abstraction mechanisms similar to those of semantic data models they display serious shortcomings in the ability to model both the dynamic nature and the many-faceted nature of common real-word entities. In this paper, formal relationships between classes of objects based on semantic data models (i.e., conceptual classes) and classes of objects based on object models (i.e., system classes), are established. The proposed method starts from conceptual classes and generates system classes. This method can be employed to transform a conceptual schema into a system schema, therefore using currently available objectoriented database systems. As a consequence, the database applications can be specified with the flexibility of a conceptual schema while the obtained implementation exploits the efficiency of an objectoriented database system. Both the correctness and the completeness of our method are demonstrated and a discussion on its complexity is provided.