The purpose of this paper is to compare different features of two approaches to the problem of tracking-prescribed trajectories in a nonlinear system, and to propose a possible merger of these approaches in what is called here the problem of "Asymptotically Robust Perfect Tracking". The latter is the problem of designing a feedforward/error-feedback control law achieving perfect tracking in case the initial conditions of the system/exosystem are known and compatible with perfect tracking, and asymptotic tracking in all other cases.