Interoperability and object identity

Abstract

Data model transparency can be achieved by providing a canonical language format for the definition and seamless manipulation of multiple autonomous information bases. In this paper we assume a canonical data and computational model combining the function and object-oriented paradigms. We investigate the concept of identity as a property of an object and the various ways this property is supported in existing databases, in relation to the object-oriented canonical data model. The canonical data model is the tool for combining and integrating preexisting syntactical homogeneous, but semantical heterogeneous data types into generalized unifying data types. We identify requirements for object identity in federated systems, and discuss problems of object identity and semantical object replication arising from this new abstraction level. We argue that a strong notion of identity at the federated level can only be acheived by weakening strict autonomy requirements of the component information bases. Finally we discuss various solutions to this problem that differ in their requirements with repect to giving up autonomy.