The caching technology, such as proxy servers, has been widely adopted to reduce the bandwidth and the response time for the WWW service. In active networks, nodes can execute simple codes to process the passed-through packets and can cache the passed-through data packets like a small transparent cache server. How to appropriately organize these distributed cache servers and to decide whether to cache the pass-through data or not are addressed in the paper. Simulation results show that the proposed self-organizing scheme is capable of reducing the response time up to 19% and increasing the probability that users get the data before the acceptable response time by 7%.