A mass of heterogeneous, distributed and dynamic information on the Web has resulted in “information overload”. It's an important and urgent research issue to provide users with effective information retrieval service on the Web. Web search engines attempt to solve this problem, yet their effect is far from satisfying. In this paper, a distributed and cooperative strategy for Web information retrieval is proposed to substitute the centralized mode adopted by the current search engines. Then a new information retrieval system model (IRSM) is presented, which supports the retrieval of metadata about Web documents and uses Z39.50 standard protocol to unify the heterogeneous interfaces of different systems. Based on them, a distributed and cooperative information retrieval framework (DCIRF) is designed to help users search the Web effectively.