Anticipating unanticipated tool interoperability using role models

The interoperability of tools heavily relies on their ability to exchange shared data. While the definition of standardised metamodelling languages such as the Essential Meta Object Facility (EMOF) [23] has substantially simplified the task of reading and persisting arbitrary domain data, there are still open issues concerning the integration of domain abstractions (metamodels) used by different tools. For example, accessing common data by shared metamodels is limited, because of the lack of first-class support for metamodel composition. Data that is processed using multiple tools must be either stored in a common abstraction---which introduces a strong coupling of the involved tools---or is replicated (e.g., represented in different tool formats)---which introduces the need for tedious synchronisation. In this paper we present how role-based metamodelling can overcome these limitations and provide a formalism to enable tool interoperability by role composition. Based on a running example, the implications of the current problems of tool integration are shown and their resolution based on role modelling is discussed.