Type Refinements

Despite many concentrated research efforts in various areas such as software engineering, programming languages, and logic, software today is not fundamentally more reliable than it was a decade ago. Software is becoming increasingly complex and inter-reliant and the techniques and tools provided by the academic community are used only sparsely. In part, this can be attributed to the many barriers to technology transfer. However, one can also recognize that in a number of ways the methods provided by the research community fail to be applicable to the problems faced by developers or maintainers of large-scale, long-lived systems. One important aspect of software development and maintenance is to understand properties of a complete system, its individual components, and how they interact. There is a wide range of properties of interest, some concerned only with the input/output behavior of functions, others concerned with concurrency or real-time requirements of processes. Upon examining the techniques for formally specifying, understanding, and verifying program behavior available today, one notices that they are almost bi-polar. On the one extreme we find work on proving the correctness of programs, on the other we find type systems for programming languages. Both of these have clear shortcomings: program proving is very expensive, time-consuming, and often infeasible, while present type systems support only minimal consistency properties of programs. The proposed research is intended to help bridge this gap by designing and implementing more refined type systems that allow rich classes of program properties to be expressed, yet still be automatically verified. Through careful, logically motivated design we hope to combine the best ideas from abstract interpretation, automated program analysis, type theory, and verification. In the remainder of this section we explain and justify our approach in somewhat more detail, before giving a research plan in the next section.