Modern digital libraries require user-friendly and yet responsive access to the rapidly growing, heterogeneous, and distributed collection of information sources. However, the increasing volume and diversity of digital informationavailable online have led to a growing problem that conventional data management systems do not have, namely nding which information sources out of many candidate choices are the most relevant to answer a given user query. We refer to this problem as the query routing problem. We introduce in this paper the notation and issues of query routing, and present a practical solution for designing a scalable query routing system based on multi-levelprogressive pruning strategies. A key idea is to create and maintain user query pro les and source capability pro les independently, and to provide algorithms that can dynamically discover relevant information sources for a given query through the smart use of user query pro les and source capability pro les, including th...