A dynamic simulator using constraint-based method is proposed. It is the extension of the formalism previously introduced by Ruspini and Khatib by including static and dynamic friction without friction cone discretization. The main contribution of the paper is in efficiently combining the operational space formulation of the multi-body dynamics in the contact space and solving for contact forces, including friction, using an iterative Gauss-Seidel approach. Comparing to existing work in this domain, we illustrate our method with scenarios involving humanoid in manipulation tasks while contacting with the environment; an experiment validates our results. Technical details that allow an efficient implementation and problems with future orientation to improve the simulator are also discussed. This work is aiming to be a potential module of the next OpenHRP simulator generation.