A method of optimizing data streaming in a peer-to-peer architecture that comprises a plurality of clients in a chain, a peer-to-peer data streaming system having such architecture, and a client terminal for use in that system. Each client monitors its own bandwidth, informs a succeeding client in the chain of that bandwidth, compares its own bandwidth with the bandwidth of a preceding client in the chain and, in response to a difference between the compared bandwidths, reorders its position among the clients in the chain. The chain thus dynamically self-organizes itself to stream data more efficiently and with higher, more reliable throughput, reducing the processing power necessary to stream the data and enabling higher quality to be achieved within the existing internet infrastructure. This also solves the `bottle-neck` problem within the cascaded streaming path by continuously organizing the participating terminals into the most efficient configuration, without interrupting the streamed data.