The goal of this paper is to present a set of design principles for avoiding replay attacks in cryptographic protocols. The principles are easily applied to real protocols and they do not consume excessive computing power or communications bandwidth. In particular, we describe how to type-tag messages with unique cryptographic functions, how to inexpensively implement the full information principle with hashes, and how to produce unique session keys without assuming mutual trust between the principals. The techniques do not guarantee security of protocols, but they are concrete ways for improving the robustness of the protocol design with relatively low cost.