Peer-to-Peer vs. the Internet: A Discussion on the Proper and Practical Location of Functionality

Peer-to-peer information sharing has become one of the dominant Internet applications, measured not only in the number of users, but also in the network bandwidth consumed. Thus, it is reasonable to examine the location of support functionality such as self-organisation, resource discovery, multipoint-to-multipoint group communication, forwarding, and routing, to provide the needed service to applications while optimising resource usage in the network. This position paper is intended to stimulate discussion in two related areas: First, where {\em should} functionality to support peer-to-peer applications be located: in the network, or as an application overlay among end systems. Second, where {\em can} functionality be located, given the practical constraints of the modern Internet including closed systems and middleboxes, as well as administrative, legal, and social issues. We will discuss the performance implications of these decisions, including whether low latency bounds for delay sensitive peer-to-peer applications (such as distributed network computing) can ever be achieved in this environment.