Epidemic, or probabilistic, multicast protocols have emerged as a viable mechanism to circumvent the scalability problems of reliable multicast protocols. However, most existing epidemic approaches use connectionless transport protocols to exchange messages and rely on the intrinsic robustness of the epidemic dissemination to mask network omissions. Unfortunately, such an approach is not networkfriendly, since the epidemic protocol makes no effort to reduce the load imposed on the network when the system is congested.In this paper, we propose a novel epidemic protocol whose main characteristic is to be network-friendly. This property is achieved by relying on connection-oriented transport connections, such as TCP/IP, to support the communication among peers. Since during congestion messages accumulate in the border of the network, the protocol uses an innovative buffer management scheme, that combines different selection techniques to discard messages upon overflow. This technique improves the quality of the information delivered to the application during periods of network congestion. The protocol has been implemented and the benefits of the approach are illustrated using a combination of experimental and simulation results.