Hosagrahar Visvesvaraya Jagadish

Managing Conflicts between Rules

Rules are used as a programming paradigm in several application domains, including active databases, planning, expert systems, and billing. For example, active databases have rules that execute upon the occurrence of particular events if specified condition predicates are satisfied. It is often the case that multiple rules are fireable when a particular event occurs. We propose a declarative mechanism to control the interaction and execution of multiple rules. The mechanism is based upon logical meta-rules that can express various types of relationships between rules. The meta-rules allow us to reason statically about the rule behavior. We can determine, in polynomial time, whether a rule will never execute, whether two rules can ever be executed together, and whether a rule system is guaranteed to have a unique execution set for all possible rules that become fireable. In this paper, we illustrate our techniques using rules in an active database. A system based upon the meta-rules and the static analysis presented here has been found to be of value in a billing application at AT & T to control interactions between discount plans.

Managing conflicts between rules (extended abstract)

Rules are used as a programming paradigm in several application domains, including active databases, planning, expert systems, and billing. For example, active databases have rules that execute upon the occurrence of particular events if specified condition predicates are satisfied. It is often the case that multiple rules are fireable when a particular event occurs. We propose a declarative mechanism to control the interaction and execution of multiple rules. The mechanism is based upon logical meta-rules that can express various types of relationships between rules. The meta-rules allow us to reason statically about the rule behavior. We can determine, in polynomial time, if a rule will never execute, whether two rules can ever be executed together, and whether a rule system is guaranteed to have a unique execution set for all possible rules that become fireable. In this paper, we illustrate our techniques using rules in an active database. A system based upon the metarules and the static analysis presented here has been found to be of value in a billing application at AT&T to control interactions between discount plans, and is presently being implemented within the application.