Consider a continuous query where a user wants to be informed when the net worth of his/her stock portfolio changes by more than a specified threshold. In this paper we develop a data dissemination technique for the Web where (a) such queries access data from multiple sources and (b) the HTTP protocol - which is inherently pull based - is used for accessing the sources. Key challenges in supporting such queries - which also arise in a diverse set of contexts including monitoring of patients, network traffic, and experiments - lie in meeting users' consistency requirements while minimizing network and server overheads, without the loss of fidelity in the responses provided to users. We also show the superior performance of our technique when compared to alternatives based on periodic independent polling of the sources.