This work is a synthesis of our experience over parallel kinematic machine control, which aims at changing the standard conceptual approach to this problem. Indeed, since the task space, the state space and the measurement space can coincide in this class of mechanism, we came to redefine the complete modeling, identification and control methodology. Thus, it is shown in this paper that, generically and with the help of sensor-based control, this methodology does not require any joint measurement, thus opening a path to simplified mechanical design and reducing the number of kinematic parameters to identify. This novel approach was validated on the reference parallel kinematic mechanism (the Gough-Stewart platform) with vision as the exteroceptive sensor.