In control design, a particular method consists in two main steps. Firstly, a global model is defined by embodying both, the system to be controlled and an exosystem (regrouping reference and disturbances models). Secondly, a standard H2 (or H1) optimization problem is defined. Formerly named the Disturbance Modeling method in the LQG context, the control goal is to minimize a quadratic index by weighting the output reference error in one side, and the control input in the other. In the case where the exosystem is unstable because e.g. persistent disturbances, this index is not bounded. It is then possible to change the criterion, weighting the ad hoc control input and output deviation to make the problem well posed (comprehensively stabilizable). The current paper deals with the same framework for stable exosystems and unstable ones, and proposes a systematic way to get a well-posed control optimization problem leading to a 2 Degree of Freedom controller.