This paper describes a strategy for the construction of large-scale heterogeneous sys- tems which combines the benefits of both tight-coupling systems (insulating users from the complexity of semantic heterogeneity thus promoting greater usability) and loose-coupling systems (allowing changes in individual components to be contained locally thus assuring the long-term viability of the system as a whole). We define a new data model (COIN) based on a deductive and object-oriented language with which knowledge of data semantics is rep- resented in the form of enriched schemas, contexts and underlying domain models. Query mediation in this framework is driven by a top-down evaluation strategy which yields a query plan describing what sources are used and what conversions must be applied in mediating semantic conflicts. The COIN data model induces a novel dichotomy between schemas and contexts thus permitting knowledge of data semantics to be shared and reused across dif- ferent systems in similar environments. This in turn provides various options for querying disparate sources, with each offering a different level of transparency depending on the needs of the users. The richness of the model offers receivers more flexibility, both in defining what disparities should count as conflicts, and how conflicts should be resolved.

The Context Interchange strategy has been proposed as an approach for achieving in- teroperability among heterogeneous and autonomous data sources and receivers (25). We have suggested (10) that this strategy has many advantages over traditional loose- and tight- coupling approaches. In this paper, we present an underlying theory describing how those features can be realized by showing (1) how domain and context specific knowledge can be represented and organized for maximal sharing; and (2) how these bodies of knowledge can be used to facilitate the detection and resolution of semantic conflicts between different sys- tems. Within this framework, ontologies exist as conceptualizations of particular domains and contexts as "idiosyncratic" constraints on these shared conceptualizations. In adopting a clausal representation for ontologies and contexts, we show that these have an elegant logical interpretation which provides a unifying framework for context mediation: i.e., the detection and resolution of semantic conflicts. The practicality of this approach is exempli- fied through a description of a prototype implementation of a context interchange system which takes advantage of an existing information infrastructure (the World Wide Web) for achieving integration among multiple autonomous data sources.