Linear Constraints Clause Samples

Linear Constraints. Now that we have addressed the motivation and interpre- tation for policies regarding acceptable adaptlet status, we turn to the matter of adaptlet parameter constraints. Every adaptlet may need to be configured according to a set of parameters. This is analogous to Component Oriented Programming [16]. For example, in the Java Beans com- ponent model every bean may expose a set of attributes for deployment time configuration. However, in our scenario we must allow for joint agreement, between client and server, of the session-time parameters. For this purpose, we allow the representation of linear con- straints [15, 4] over adaptlet parameters. For example, a linear constraint could be used with two adaptlets imple- ▇▇▇▇▇▇▇ a service-level agreement, −2.0 ∗ PayFeature.price + 1.0 ∗ QoS.guarantee = −100.0 This constraint sets the price of a connection at fifty dollars plus half the amount of bandwidth reservation. Graphically, this allows clients and servers to negotiate a choice of the two aspect parameters (referenced as fields of the aspects) anywhere along the line defined by the equation. In contrast to systems of non-linear constraints, linear sys- tems are decidable and tractable. Thus far we have not explored support for non-linear constraints. Modern solvers are usually based on the Simplex [4] algorithm due to Dantzig.