Mathematics of Bandwidth Sharing and the Future Internet F. Baccelli. * Nobody has ever produced a comprehensive quantitative model for the dynamics of the competition between flows that takes into account 1) the congestion control algorithms used or to be used in the Internet; 2) the statistical structure of typical Internet applications; 3) the basic queueing/scheduling mechanisms used in access/core networks. The design of new distributed congestion control algorithms with predictable statistical behavior will require some mathematical breakthrough on the modeling of flow competition within this context. * The dynamics of large overlay networks can be seen as yet another type of interaction between a large collection of point to point flows, which involves competition, but also collaboration (e.g. client-server, split etc.), with as a corollary new phenomena such as starvation, overflow at overlay nodes etc. The understanding of the interplay between these local point to point mechanisms, the overlay control and classical applications is a major challenge. * From a given geographic point, the Internet will more and more be accessed through a variety of coexisting wireless technologies: WIFI meshes, WIMAX, 3G/LTE, cellular etc. Current routing and congestion control mechanisms succeed very well in sharing dynamically the wired resources of the Internet in a fully distributed way. There is no analogue of this for efficient spectrum sharing in heterogeneous wireless access networks yet. There is a need for scalable user association/split algorithms, power control algorithms, access scheduling algorithms and congestion control mechanisms within this context that take into account the specificities of wireless channels. For addressing these questions, new mathematical tools will be needed from such disciplines as probability theory, statistics, control, discrete event dynamical systems, statistical physics. Scalable simulation methodologies will also be essential within this framework.