Federative Databases

It seems that the need for considering distributed data bases rather than a central one is motivated by:

• the role played by data bases in Artificial Intelligence when an expert system requires volume processing of data and rules. In the specific case of a distributed expert system (see above), the distribution of data/rules requires to consider the distribution of the data base and the elaboration of a Cooperating Knowledge Based System (CKBS),
• for some complex domains of applications (like, e.g., Computer Integrated Manufacturing applications [Kem90]), the data base system is inherently distributed.
• More and more applications are designed and needed that integrate and use data and services of local (data base) systems [Bro92]. Information has become a key factor in industry and it becomes essential to establish relationships between already existing resources and service (e.g. by coupling administrative and production systems, both equipped with local data bases).

In all cases, the specification of a federative system presents strong analogies with the specification of composite systems [Bro92][Buc91]. In [SJH93][Jun90][Deen90], such concerns motivate the identification of `object' (this concept should not be confused with the concept of object in object-oriented programming), each of them incorporating some knowledge about an application domain (like facts, rules, strategies, etc) and being a unit of design that can be composed with other objects via aggregation and specialization hierarchies. Within this framework, particular attention is devoted to:

• abstract application modeling: integrate distributed and heterogeneous resources through global applications [SWS91];
• remodeling structures and services of local systems in terms of a common abstract framework [Buc91]: make local systems compatible and resolve semantic incompatibilities and overlaps [SJ92][SJH93];
• autonomy and interfaces: it is still not clear to which extent local data bases can be autonomous and how access can be controlled; one promising way is through views [KDN90];
• cooperation strategies: different architectures of distributed data bases exist and provide solutions for different cooperation strategies meeting different objectives [WR92][Dur89];
• reliability and recovery facilities with specific properties related to an individual data base viewed as an agent [Wie89].

• Experts of the group: