We consider reliable multicast in distributed systems including mobile hosts (MHs) that communicate with a wired infrastructure by means of wireless links. Nearly all existing proposals are based on hand-off, i.e. whenever a MH switches cell, state information about this host travels across the wired network from the support station of the old cell to that of the new cell. However, we are not aware of any detailed performance analysis for hand-off based reliable multicast protocols: previous research in this area has focused mainly on correctness rather than on performance. We analyze in detail, by simulation, the performance of a proposal by Acharya and Badrinath that is based on hand-off and has been highly influential in the design of later protocols. Then, we compare this proposal with one by us that is based on an entirely different philosophy and is the only existing proposal not based on hand-off. Surprisingly, we found that our proposal outperforms the one by Acharya and Badrinath in all the aspects considered: latency, scalability, bandwidth usage efficiency and quickness in managing cell switches of MHs. Moreover, we found that this performance improvement is not obtained at the expense of increased resource requirements on MHs such as energy or memory. We believe that this performance and cost analysis allows us to gain insights into the design of reliable multicast protocols for distributed mobile systems.