Formal Semantics for Composite Temporal Events in Active Dbs 3725

He is currently a consultant with Cambridge Technology Partners, where he specializes in data warehouse and data mining applications. He was previouslywith NCR Teradata. His current research interests also include active and temporal reasoning in databases. Dr. Zaniolo's research interests include database systems, non-monotonic reasoning in databases and knowledge bases, and document servers for the informationage. He served as program chairman of several conferences, including ACM SIGMOD and VLDB conferences. 35 rules, whereby they can be implemented incrementally in a history-less manner 7]. This leads to implementation techniques similar to those used for supporting the class of XY-stratiied programs in LDL++. Forward-chaining production systems of more procedural nature, such as CLIPS can also provide a sound basis for implementation, due to their eecient rule activation mechanisms. The approach proposed in this paper lays the seeds for further research. We have already studied the issue of temporal aggregation in active rules and the results can be found in 25]. An important issue yet to be answered is the formal comparison of diierent composite event speciication languages. In this respect, Datalog 1S provides a sound formal basis, due to the fact that its formal semantics is well-understood and its expressive power w.r.t. to other languages (temporal or otherwise) has been previously characterized 3]. Within this framework, the expressive power of such languages can now be studied using methods and tools previously used for Datalog and First-Order languages 1]. Another topic for further research that we are currently pursuing is that of alternative optimization strategies for composite event languages. For instance, while we have implicitly assumed that Datalog 1S rules are evaluated incrementally as new basic events occur, composite events can also be evaluated and optimized as queries against a historical database of events. Acknowledgments The authors would like to thank Reza Sadri and the referees for their suggestions. Notes 1. The syntax used for conditions is in fact that of the where clause in SQL; however, in EPL, the comma can also be used to denote the logical AND. 2. Note i that the Timestamp parameter is supplied by the database system, and thus, it is treated as a data parameter. 3. From now on, unless otherwise indicated, variables will denote tuples. sequence, the conjunction and the disjunction constructs. 5. We use the binary construct *seq in place of the * EPL construct, so that the representation is more compact and easier …