This work is concerned with scaling decentralized content-based publish/subscribe (CBPS) networks for Internet-wide content distribution. A fundamental step for CBPS networks to reach the Internet-scale is to move from the exhaustive filtering service model, where a subscription selects every relevant publication, to a service model capturing the quantitative and qualitative heterogeneity of information consumers' requirements. In previous work, we described a service model allowing information consumers to express the maximum number of publications they would like to receive per service period and how to take advantage of such knowledge to pace the dissemination process. This paper extends by introducing a generic service model that seamlessly supports content-based information retrieval and dissemination and investigates through extensive simulations the performances of six caching policies in terms of consumers satisfaction and bandwidth usage.