We investigate the fairness and throughput properties of a simple distributed scheduling policy, maximal scheduling, in the context of a general ad-hoc wireless network. We design a fully distributed algorithm that combines a token generation scheme with maximal scheduling policy so as to attain maxmin fair rates within the feasible region of maximal scheduling. We next present throughput guarantees of maximal scheduling that quantify the performance loss of each session due to the use of local information based scheduling. We show that the performance loss for each session depends on the maximum “interference degree” in its neighborhood. We also demonstrate that the performance penalties can not be localized any further