This paper proposes a simple data structure, called a prefix list, which maintains all prefixes of a string in reverse lexicographic order. It can be on-line incrementally constructed in time and space linear in the string length. It is strongly related to sufix trees and sufix arrays, and may share applications with these existing structures. A sufix array can be built via the corresponding prefix list in linear time. Particular applications of the prefix list lie in source-coding problems that require on-line right-to-left string matching. We apply the prefix list to on-line estimation of source entropy and to context-based symbol-ranking text compression algorithms.