Clarify and Efficiency in Design

Award Abstract 0613913 PIs Yanhong A. Liu and Scott D. Stoller SUNY Stony Brook TITLE SoD: Clarity and Efficiency Abstraction for design, such as object abstraction, supports the separation of what operations are provided by systems and components from how the operations are implemented, and is essential in the construction of complex systems from components. Unfortunately, clear and modular designs have poor performance when expensive query operations are performed, while efficient designs that properly maintain the query results are much more difficult to develop and to understand, because the design may blow up significantly and is no longer clear or modular. This project proposes to create a general framework and language for expressing design knowledge that support both clarity and efficiency, by rigorously capturing invariants, ways to efficiently and incrementally maintain invariants, and the associated costs; and to develop a method for automatically applying the design knowledge. The underlying principles of incremental computation are the counterpart of differentiation in calculus and will help lay a solid foundation for science of design. This project also proposes to implement the framework and method, apply them to important application domains, including security policy frameworks and operating systems, and teach them to both undergraduate and graduate students.