Addresses the design of the reengineering service for the DOK (Distributed Object Kernel) federated database. This service allows the hiding of the heterogeneity of databases involved in a federation by generating object-oriented representations from their corresponding schemata. We propose a complete methodology that supports the identification and the translation of both the explicit and implicit information. The identification of object-oriented constructs is performed by classifying a relational schema into different categories of relations, namely base, dependent and composite relations. The main difficulty in designing the reengineering service relies on the distinction between the different types of relationships amongst classes. Our approach deals with this problem by analysing relations according two types of correlation: (i) the degree of correlation between the external and primary keys, and (ii) the degree of correlation between sets of tuples in the relations. Examining these correlations uncovers implicit relationships contained as well-hidden classes in a relational schema.

We address the development of a normalization theory for object-oriented data models that have common features to support objects. We first provide an extension of functional dependencies to cope with the richer semantics of relationships between objects, called path dependency, local dependency, and global dependency constraints. Using these dependency constraints, we provide normal forms for object-oriented data models based on the notions of user interpretation (user-specified dependency constraints) and object model . In constrast to conventional data models in which a normalized object has a unique interpretation, in object-oriented data models, an object may have many multiple interpretations that form the model for that object. An object will then be in a normal form if and only if the user's interpretation is derivable from the model of the object. Our normalization process is by nature iiterative, in which objects are restructured until their models reflect the user's interpretation.