Today's Internet's routing paths are inefficient with respect to both connectivity and the market for interconnection. The former manifests itself via needlessly long paths, de-peering, etc. The latter arises because of a primitive market structure that results in unfulfilled demand and unused capacity. Today's networksmake pairwise, myopic interconnection decisions based on business considerations that may not mirror considerations of the edge networks (or end systems) that would benefit from the existence of a particular interconnection. These bilateral contracts are also complex and difficult to enforce. This paper proposes MINT, a market structure and routing protocol suite that facilitates the sale and purchase of end-to-end Internet paths. We present MINT's structure, explain how it improves connectivity and market efficiency, explore the types of connectivity that might be exchanged (vs. today's "best effort" connectivity), and argue that MINT's deployment is beneficial to both stub networks and transit providers. We discuss research challenges, including the design both of the protocol that maintains information about connectivity and of the market clearing algorithms. Our preliminary evaluation shows that such a market quickly reaches equilibrium and exhibits price stability.