The decentralized control problem for linear dynamic systems is revisited using a parameter space approach which enables the definition of the decentralized feedbacks from the existence of non-empty parameter convex sets. The convexity property enables the derivation of efficient numerical algorithms based on standard approaches in convex programming. The continuous-time and discrete-time cases are investigated and the decentralized control design is also treated to meet other important assignments such as : optimal H-2 performance index, absolute stability, H-infinity, prescribed attenuation and robustness against actuator failures. Some numerical experiments illustrate the potential of this new control design.