Null Values Revisited in Prospect of Data Integration

A considerable part of the information available in the real world is inherently imperfect and/or incomplete. Traditionally, the most commonly adopted modelling approaches for dealing with imperfect and missing information are based on the use of default values and of null values. However, in order to deal with imperfect information in a more efficient way, a database system needs some more advanced modelling facilities that better reflect the semantics of the data. Such facilities become even more necessary if data stemming from different data sources must be integrated in a distributed, federated database system. In this paper, the concept of a ‘null’ value has been revisited. A semantically richer definition, based on possibility theory, is proposed together with a description of the accompanying many-valued logic. Additionally, the potentials of the new approach with respect to the integration of data in multi-database environments or grid database services are illustrated.