In infrastructure networks the AP stays on the same network all the time. This property ensures that packets transmitted from cards connected to an infrastructure network always get through, and packets to these cards are buffered at the APs if the card is currently not in that network (or sleeping). The performance of the network is affected only by the buffering capability of the AP and the switching card. It is unaffected by the number of switching cards. For all the cases that we studied, the end-user performance is always good with packets sent by or to the switching cards always getting through.
The situation is different for ad hoc networks, and we describe the problem using the following scenario. Suppose Bob and Alice are co-workers who belong to the same organization. Further, suppose Bob is sharing a presentation with Alice over an ad hoc network. Both are also connected to the corporate infrastructure network using MultiNet. A worst case arises when both Bob and Alice switch synchronously to the other network. That is, Bob switches to the infrastructure network just when Alice switches to the ad hoc network, and vice versa. As a result Bob and Alice are never in the same network at the same time, and therefore will never be able to communicate. Note that Bob and Alice might be able to communicate over the ad hoc network if there is a another person, say Trudy, within the range of both Alice and Bob, whose time in the ad hoc network overlaps with the duration of both Bob and Alice. However, there might be situations where Trudy is not present. Our switching protocol is designed to handle such scenarios.