We discuss control strategies for communication networks such as the Internet. We advocate the goal of welfare maximization as a paradigm for network resource allocation. We explore the application of this paradigm to the case of parallel network paths. We show that welfare maximization requires active balancing across paths by data sources, and potentially requires implementation of novel transport protocols. However, the only requirement from the underlying `network layer’ is to expose the marginal congestion cost of network paths to the `transport layer’. We further illustrate the versatility of the corresponding layered architecture by describing transport protocols with the following properties: they achieve welfare maximization when each communication may use an arbitrary collection of paths, available in an overlay, combined in series and parallel.We conclude by commenting on incentives, pricing and open problems.