Although caching has been shown as an efficient technique to reduce the delay in generating web pages to meet the page requests from web users, it becomes less effective if the pages are dynamic and contain dynamic contents. In this paper, instead of using caching, we study the effectiveness of using pre-fetching to resolve the problems in handling dynamic web pages. Pre-fetching is a proactive caching scheme since a page is cached before the receipt of any page request for the page. In addition to the problem of which pages to be pre-fetched, another equally important question is when to perform the pre-fetching. To resolve the prediction and timing problems, we explore the temporal properties of the dynamic web pages and the timing issues in accessing the pages to determine which pages to be pre-fetched and the best time to pre-fetch the pages to maximize the cache hit probability of the pre-fetched page. If the required pages can be found in the cache validly, the response times of the requests can be greatly reduced. The proposed scheme is called temporal pre-fetching (TPF) in which we prioritize pre-fetching requests based on the predicted usability of the to-be pre-fetched pages. To minimize the impact of incorrect prediction in pre-fetching on processing of on-demand page requests, a qualifying examination is performed to remove unnecessary and low usability pre-fetching requests while they are waiting to be processed and just before their processing. We have implemented the proposed TPF scheme in a web server system and experiments have been performed to study its performance characteristics compared with conventional cache-only scheme using a benchmark auction application under different system and application settings. As shown in the experiment results, the overall system performance, i.e., response time, is improved as more page requests can be served immediately using pre-fetched pages.