Gossip-based protocols are now acknowledged as a sound basis to implement collaborative high-bandwidth content dissemination: content location is disseminated through gossip, the actual contents being subsequently pulled. Despite their success, gossip-based systems do not constitute the panacea for all the problems in large-scale distributed computing. The majority of existing protocols, implicitly assume that all nodes are created equal. Nodes usually have the same number of gossip targets (fanout) and the same dissemination period, and therefore send exactly the same number of messages. However, this uniform distribution of load ignores the heterogeneous nature of large-scale distributed systems where nodes frequently exhibit significant differences in their capabilities. In this paper, we first examine how heterogeneous capabilities such as available upload bandwidth may influence the behavior of gossip-based protocols. Then we outline possible solutions ranging from adapting the fanout of nodes, to structuring the dissemination process.