Peer-to-peer technology allowed massive free music file sharing, which led to much controversy, legal debate, and a great deal of speculation about the technology's business viability. Operation of a peer-to-peer (P2P) network depends on the content and bandwidth contributions from its users. Many users free ride by contributing fewer resources than they use. Research found that free riding intensifies with crowding. These salient features suggest that the underlying template describing a P2P system may be the limits to growth archetype. We expand on this idea by building a system dynamics model, which captures complex causal interactions between private provision of common resources, free riding, network performance, and membership dynamics. Base runs confirm that the model reproduces historically observed behavior. We use the calibrated model to test effectiveness of certain modifications that have been proposed for real P2P networks. Simulation results are explained by reviewing the interplay of reinforcing and limiting feedbacks that drive the system. Copyright © 2004 John Wiley & Sons, Ltd. Syst. Dyn. Rev. 20, 237-262, (2004 In 1999, 19-year old Shawn Fanning released Napster software. Its sole purpose was to allow music lovers to swap free MP3 files. Within 20 months, 65 million people used the software to trade music (Leuf 2002, p. 191). A group of big record labels responded with a fierce legal offensive against Napster, Inc. The group claimed that the P2P network violated copyright laws and cost them millions of dollars in lost CD sales (CNN Money 2002). Napster, Inc. eventually lost the case in court and had to shut down. However, a number of peer-to-peer networks withstood legal challenges and are thriving. Two popular systems are Gnutella and KaZaA. Unlike Napster, current systems do not require a central server or a centralized controlling body.The phenomenal popularity of the technology led to the predictions of an approaching peer-to-peer revolution in electronic commerce. Commercial adoption of P2P systems may be driven by its many coveted features: distributed cost of storage, distributed processing cost, autonomy of nodes, robustness against malicious attacks, and load balancing (Yang and Garcia-Molina 2002a About 60 per cent of Americans, or 174.6 million people, regularly use the Internet (Lenhart et al. 2003). Millions of additional devices including various appliances will come online after the adoption of the IPv6 protocol (Tanenbaum 2003, p. 464). The pressing issue is whether or not peer technology can scale to accommodate all the potential online users. Some experts have concluded that peer-to-peer technology is not capable of handling large volumes of traffic required for successful scaling (Ritter 2001). Proponents of the technology disagreed (Kabanov 2001), suggesting that as nodes gain access to faster connections the problem would go away (Shirky 2000).An examination of a typical peer-to-peer system suggests that it may be described by the limits to growth archetype (Sen...