The decomposition of large systems into parts is a general principle of software design. Even more, in the scope of distributed systems, a partition of the whole system into distributable components is necessary. Decisions about what constitutes a component of a system are usually either based on the behavior or on the structure of the system. Nevertheless, there is a strong mutual influence between both kinds of decomposition. Despite the importance of structural and behavioral decomposition, many modeling notations and languages define the semantics of these concepts rather vaguely, and this may lead to incorrect implementation.
