Internet rate control protocols, such as TCP, encounter severe performance problems in wireless mesh networks. Because wireless networks use shared communication channels, contention and interference can significantly degrade flow throughput and fairness. Existing research takes either an engineering-based or optimization-based approach to solve the performance problems. The engineering-based approach usually solves a specific observed problem, but does not necessarily optimize the overall performance. The optimization-based approach mathematically models the network to find the optimal resource allocation among competing flows. The model can lead to a distributed rate control algorithm with performance guarantees, but relatively little work has been done to verify that the algorithm leads to good performance in real networks. This dissertation develops a more accurate network optimization model, implements the derived distributed rate control algorithm in a mesh testbed, and discusses observations in the extensive experiments. We first synthesize models used for optimizing fair rate control for wireless mesh networks, and discuss their tradeoffs. We then propose a partial interference model which uses more accurate objective functions and constraints as compared to the binary interference model. Numerical results show that the partial interference model outperforms the binary interference model in all scenarios tested, and the results also suggest that partial interference should be modeled separately from contention. Our experimental results confirm the prevalence of partial interference in our mesh testbed, and show that the partial interference model results in significantly improved performance in a typical interference topology. We also observe a significant deviation between theory and practice, whereby, the assumption of a linear relationship between interfering links breaks in our experiments. We discuss several directions to further investigate this issue.