This paper studies the problem of congestion control on wireless networks. A dynamical model for the end-to-end network flow control that exploits the differentiation between congestion loss and physical channel error loss is proposed. The introduction of a specific wireless model is motivated by the distinctive presence of channel errors, which are often not known exactly. We assume that each wireless link is associated with an additional error function that depends on the current flow along the link and that accounts for the packet loss rate caused by the physical channel. This leads to a new dynamic flow control scheme that naturally extends a known mathematical model for the fluid-flow approximation of the Transmission Control Protocol for wireline networks. The main objective of this work is to study the dynamical properties of the new model: we analyze its nonlinear dynamics, derive its stability properties, and study its robustness to delays. We also present and discuss some ns-2 simulations of its dynamics. This work additionally looks at the actual implementation of the proposed scheme: by requiring only modifications to the application layer rather than the transport one, no alterations to the network infrastructure or transport protocols are needed. The article argues that the new scheme appears to be not only theoretically meaningful but also practically relevant for an application layer implementation. Copyright © 2011 John Wiley & Sons, Ltd.