Class-Level Behavioral Modeling and Synthesis

When modeling the behavioral requirements of objectoriented distributed systems, one has to take class-level scenarios into account. Those describe interactions that apply to all/some instances of some class. Current scenariobased notations fall short on this problem. They remain at instance level and suitable only for modeling the behavior of a particular population. State machines are often used later in the analysis to factor scenarios classwise. However, when dealing with class-level behavior, state machine models get unnecessarily awkward and implementationoriented, incorporating iterations over instances, for example. This paper brings two contributions solving this crucial issue. First, we propose an obvious extension of current scenario and state machine languages, already envisioned by Harel back in 1984: all notations get extended with universal and existential quantifiers. Second, we upgrade classical synthesis algorithms dealing with class-level notations. This extension preserves properties of instance-level algorithms. 1