Background and History

Proposals for next-generation databases [BMS84] stem from three traditions:

• Databases: This stream is best typified by extended relational and object-oriented databases. The main extension is the addition of complex domains, shared sub-objects, and procedures to the database. Examples include Postgres, Damokles, Cactis, Iris, DASDBS, and many others.
• Programming languages: The main goal is to provide persistence and sharing to one or more programming languages, ideally orthogonal to the type systems of these languages. Examples include the O system for imperative programming languages, and typed versions of Prolog such as LOGIN or PROTOS-L.
• Knowledge representation: Some of these systems aim at providing database service to AI applications, others come from the tradition of deductive databases or semantic data models and aim at formal as well as practical support for general database systems and applications. Examples include the deductive databases EKS, LDL, CORAL, NAIL!, and the concept logic systems CLASSIC, BACK, and LOOM.

A sample of the former two groups of systems is described in [TKDE90], whereas several prototypes of the latter are reported in [SIGA91]. Although a certain confluence can be observed, these systems do not only differ in their background theories but also in their intended application domains.

The first group wants to support non-standard applications such as the handling of complex engineering objects. The second group aims at an easier programming environment for standard applications: additionally, they emphasize the goal of making applications written in different languages interoperable.

The third group pursues, as one of its goals, the support for AI applications such as natural language understanding or expert systems. However, this may remain a fairly limited portion of the software market. The reason why we became interested in this group of languages is therefore a slightly different one: the important role we expect them to play in meta data management. Meta data applications range from the uniform access to heterogeneous data sources, to the coordination of design processes, to the integration of heterogeneous information services in networked production chains, whole enterprises, or even trans-national networks. In these applications which are crucial for the huge integration tasks to be tackled in the 1990s, the possibility not only to execute systems but also to reason formally about their structure and capabilities, can be considered a competitive advantage over the other approaches. ConceptBase has provided some validity to this claim by extensive usage experiments in several of the above areas.

The language supported by ConceptBase, Telos, has been one of the earliest attempts to integrate deductive and object-oriented data models ([STAN86],
[MBJK90]). The O-Telos [JEUS92] dialect supported in ConceptBase takes a conservative integration approach, with the main design goals of semantic simplicity, symmetry of deductive and object views, flexibility and extensibility. This emphasis, technically supported by a careful mapping of Telos to Datalog with negation, has paid off both in user acceptance and ease of implementation.

Development of ConceptBase started in late 1987 in the context of ESPRIT project DAIDA [Jarke, 1993] and was continued within ESPRIT Basic Research Actions Compulog 1 and 2 (1989 - 1995). Versions have been distributed for research experiments since early 1988. The previous distribution versions, ConceptBase V 3.3 and V4.1, has been installed at more than two hundred sites worldwide and is seriously used by about a dozen research projects in Europe and North America.