We study peer‚Äö?Ñ?¨to‚Äö?Ñ?¨peer (P2P) multicast streaming protocols for high bandwidth streams to large numbers of heterogeneous and transient users. First, the shortcomings of tree‚Äö?Ñ?¨based P2P schemes are discussed. Subsequently, a scalable swarm‚Äö?Ñ?¨based scheme is introduced that avoids maintaining a rigid logical topology. Peers self‚Äö?Ñ?¨organize into an unstructured overlay in an ad-hoc fashion. Stream data are divided into small‚Äö?Ñ?¨sized units and distributed in a fashion that fully utilizes the upload capacity of all peers. A credit‚Äö?Ñ?¨based incentive scheme is introduced to encourage peers to contribute their capacity. The incentive scheme operates based exclusively on local information. We evaluate the proposed scheme through simulations in a dynamic and heterogeneous environment. A key benefit of the proposed scheme is its ability to operate under resource‚Äö?Ñ?¨constrained conditions where traditional tree‚Äö?Ñ?¨based approaches typically fail.