8. Related Work

[Guyo86] J. Guyot. Un modele de traitements pour les bases de donnees: un formalisme pour la conception la validation et l'execution de la specification d'une application. [Hend86] J. Hendler. " Enhancement for multiple-inheritance ". theoretical basis for reasoning about object life-cycles. Furthermore, we have presented an algorithm which can be used for verifying consistency of specifications as well as for monitoring during run-time whether specifications are respected. PTL is clearly a suitable tool for specifying behavior evolution and coordination of objects but it has a rather steep learning curve, which means that some training is required before it can be used. Another negative point concerns consistency proofs, since, even for small formulas, such a proof may become tedious. In fact, the algorithm we presented may generate a number of nodes exponential to the number of temporal operators of the input formula. In other words, even a computer implementation of that algorithm is not a realistic solution if we consider realistic applications. Restricted forms of PTL may reduce the number of nodes of the satisfiability algorithm to polynomial size [Emer89]. It would be worthwhile to examine whether such restrictions are still suitable for our purposes. Although our ambition is to extend object-oriented models to describe dynamic aspects of applications in general, our extensions have been limited to describe behavioral evolution of objects when considered in isolation. A further step in our work, which is near completion, consists of expressing dynamic properties and constraints among interacting sets of objects, and coordinate their functionality to reach a specific goal. Data model issues for object-oriented applications ". [Chen76] P.P.S. Chen. " The entity-relationship model: Toward a unified view of data ". Objects modifying their behavior dynamically, which in our model is described by objects playing different roles, has been proposed by a number of researchers[Tsic87] [Casa88] [Pern90] [Hend86] [Arap89]. In the KNO model [Tsic87], for which a prototype system has been developed [Casa88], it is possible to specify finer-grained changes of object behavior than in our model. Objects can add or drop methods during run-time, as well as exchange methods among themselves. No means are provided to control or constrain the dynamic evolution of objects. Considerations which led the designers of KNO to include the above features in their model were the following: First, there are cases for which it is not desirable and sometimes not even possible to accurately specify all parts of …