From a language technologist's point of view, the penetration of natural language interfaces onto today's web is somewhat disappointing; it seems that information retrieval, forms based, metaphor-based and hyper-link interfaces dominate all points of the design space. While open domain question answering promises to rival or extend information retrieval systems, restricted domain question answering systems likewise represent a rival to forms-based interfaces. The purpose of this position paper is to discuss the properties of potential web-based `killer applications' of restricted domain question answering. The paper entertains a set of candidate domains, proposes a general methodology for building restricted domain interfaces and highlights some near term challenges that must be confronted.