Modeling and Implementing Knowledge-based Protocols

Many important patterns in the design and analysis of protocols can be interpreted in terms of knowledge. For example, proving that a process cannot perform a particular action at a point in time. Such proofs typically involve reasoning about points in time that a process cannot distinguish between, and these points can be used to characterize the process’ knowledge. We survey the use of knowledge in protocol design and analysis, in relation to two questions. First, we consider the question of representing what processes know. This involves describing what a single process knows, what groups of processes know, what processes know that others know and what is common knowledge. And, all this changes with time and communication. We review an influential definition of common knowledge in terms of epistemic reachability and its application to an analysis of Byzantine agreement. Second, we examine what a process can do with knowledge. Two frameworks are considered: knowledge-based programs and dynamic epistemic logics. A knowledgebased program describes a process that conditions its actions based on knowledge, e.g., “if I don’t know whether she has received the message, resend”. Knowledgebased programs are general but abstract descriptions of processes, i.e., the knowledge of a process and how it changes is determined by an externally specified or synthesized interpretation. Dynamic epistemic logics are useful in more specialized settings where a process has incomplete knowledge of a situation and acquires information via communicative acts. We review some of the positive and negative results under these paradigms, and the current trends.