Most of the existing solutions to integrate heterogeneous information systems concentrate on the definition of an enterprise- or organisation-wide data model (unique global schema) as part of an overall business process defini-tion. Based on this approach information sharing is realised through the introduction of a common database which replaces the existing application-based databases of an enterprise or organisation, and, to which all exist-ing applications have to be connected.

This so called fusion approach ignores the following constraints which are fundamental to most of existing large-scale enterprises or organisations:

• Introduction of a common database is hardly feasible, if the different information systems exist to perform very different functions; and , if there is no common management to direct and monitor the migration effort; and, if some systems are brittle legacy code that cannot or will not be changed.
• Migration of individual applications to a common database is prohibitively expensive. Further on, organi-sations have spent money to an extensive amount in database systems, in application development and in programmers know-how. Saving these investments is a pressing requirement for database integration, which, however, is less supported by the fusion approach.
• Development of a unified global schema is extremely expensive in terms of time and domain expert re-sources. Operational and developmental experts are required from each application area.
• In a dynamic organisation, individual functions are continuously evolving, as are the interactions between them. While it may be possible to capture the evolutions in pair-wise (point-to-point) interoperability mechanisms, it is difficult to track interactions between multiple evolving functions in a large system.
• In large systems, the development or integration authority may not have ownership or control of all the applications. Some may be COTS packages. Some applications may be owned (i.e. supported and paid for) by independent organisations. The integration authority may have limited control of application functional-ity, budgets, or development schedules.
• The consequence of ownership problems and functional evolution is that applications will change independ-ently of the intent or schedules of the integration authority. As the number of participating functional areas increases, the common schema cannot evolve rapidly enough to accommodate the pace and scheduling of application changes. Thus the common database may never stabilise sufficiently to provide a development baseline.
• The application business areas may have different vocabularies, levels of abstraction about system informa-tion or behaviour, and underlying business models. It may be infeasible to define a unified business model in sufficient detail to support applications for different business areas.
• In large systems, external interfaces may be defined by organisational, national or international standards over which the integration authority has no control. These external standards may not be mutually compatible.

Background and Problem Definition DMF Mission