Traditional mobile ad hoc network (MANET) routing protocols assume that contemporaneous end-to-end communication paths exist between data senders and receivers. In some mobile ad hoc networks with a sparse node population, an end-to-end communication path may break frequently or may not exist at any time. Many routing protocols have been proposed in the literature to address the problem, but few were evaluated in a realistic “opportunistic” network setting. We use simulation and contact traces (derived from logs in a production network) to evaluate and compare five existing protocols: direct-delivery, epidemic, random, PRoPHET, and Link-State, as well as our own proposed routing protocol. We show that the direct delivery and epidemic routing protocols suffer either low delivery ratio or high resource usage, and other protocols make tradeoffs between delivery ratio and resource usage.