How to extend the life of a battery powering a mobile communications device, such as a personal communications system (PCS) station or cellular phone, is of great relevance to users, equipment manufacturers, and service providers. Though the subject may not be of primary interest to readers of *Computing Reviews,* some parts of this paper are closely related to computer science, and I will focus on those.

The paper deals with the following problem: A mobile station wants to transmit data. There is some external interference near the transmitter. This interference may cause reception errors, unless the signal-to-interference (SIR) ratio is high enough. This interference has a cyclic pattern, which does not change over relatively long periods of time. Interference near the receiver is not considered in this scenario. The authors’ goal is to devise an adaptive algorithm that will determine when, and at what power, the transmitter should transmit to achieve a given statistical bit error rate.

They present an algorithm that does precisely that. First, they build a mathematical model, which yields a system of integral inequalities to be minimized. The algorithm is derived from the model by substituting sampling for the unknown interference function.

Finally, the quality of this algorithm (defined in terms of the average power used in a transmission) is evaluated by simulation and compared with the power used if a constant SIR is maintained. Not unexpectedly, the adaptive scheme performs substantially better than maintaining a fixed SIR.

The paper meets the authors’ objectives. From the perspective of a computer scientist, however, both the presentation of the adaptive algorithm and the simulation experiments are flawed. The algorithm is presented in a way not seen since the early 1970s; among other things, 8 out of 13 steps of the “sub-algorithm” end with *goto.*

The discussion of the simulation omits the initial value of the variable *step,* even though it clearly affects the accuracy of the results. Also, the simulation time is unusually brief. In a mobile environment, the nature of interference changes too rapidly to warrant simulating a fixed interference source for longer periods. This, however, brings to mind a methodological doubt. Everyone will agree that adaptive algorithms are better when the conditions to which they are adapting remain constant, but what about their behavior in transition periods? It would be interesting to see how the authors’ adaptive algorithm performs in such conditions.