Transformational design of distributed systems

Summary form only given. Our position is founded upon three premises. The first is that transformational techniques can provide useful support to the design of distributed systems. In this approach, design commences with an idealized specification (expressed in a non-distributed fashion), and terminates with a distributed implementation that exhibits the same, or acceptably close, functionality as the specification. This leads to the second premise, the need to employ imperfect transformations that potentially compromise or approximate the specification in the course of design; in this respect we deviate from the mainstream approach to program transformation. Furthermore, distributed systems are often fielded in a highly variable and unpredictable environment. This, combined with the use of imperfect transformations, motivates the third premise, that distributed systems must often be self-monitoring, ie., that the system monitor its behavior and performance during its interactions with its environment.<
>