SUMMARYThin client computing trades local processing for network bandwidth by off-loading application logic to remote servers. User input and display updates are exchanged between client and server through a thin client protocol. In a wireless device context, it is important to achieve bandwidth efficient thin client protocols because bandwidth availability is limited and the power consumption of the wireless network interface card is directly related to the amount of data that is sent and received. This paper presents and evaluates a novel client-based mechanism which is transparent to the server to reduce upstream bandwidth consumption. Typically, thin client protocols encode user input as a series of small packets, resulting in a major packetization overhead. By buffering user events at the thin client protocol layer, this overhead can be reduced. However, buffering strategies might result in increased response delays for the user. Therefore, models of the upstream bandwidth and the user perceived responsiveness of pull thin client protocols are presented and validated. These models are used in an adaptive framework, which determines the appropriate buffering time to minimize the bandwidth as much as possible without degrading the responsiveness. For lower network roundtrip times and users actively generating input, it is shown how bandwidth savings up to 78% can be achieved while keeping the average user perceived responsiveness below 150 ms.