In the present web search, mining the web repository and yielding the most relevant result set that is minimal in size is the most sought after requirement. In this paper we proposed a new protocol namely User Relevancy Improvisation Protocol (URIP) that employs "push" mechanism to obtain the most current and concise summary of the web sites. We strongly feel that context identification and relevancy characterization are best achieved through collaboration between web servers and search engines. Relevancy being user specific, we introduced two parameters, Knowledge Quotient (Kq) and Knowledge Factor (Kf) to rank the web pages. URIP clients at the respective web servers compute these parameters for each web page. Search Engine Server embodies the URIP server that gets the update notifications from all the web servers as and when the web sites are created or modified. User input contains query terms, his/her knowledge level (Kq́) and domain expertise (Kf). Relevant document set is retrieved with closely matching Kq, Kq' and Kf, Kf. Apparent advantages of our proposal are, (1) Does not need constellation of thousands of systems to index the web data, (2) Search engine site need not maintain a snapshot of the web to respond to the users, (3) Eliminates spider activity and most importantly (4) Yields minimal and most relevant result set. We compare our results with the popular search engine Google. A Finite State Machine (FSM) based specification for the protocol is given and the protocol is analyzed using M/G/1 model. © Springer-Verlag 2003.