With the emergence of more and more torque-controlled robots, the agility of leg locomotion has been promoted to a new level, which, however, seems also to imply that only torque-controlled robots are appropriate for leg locomotion. This paper demonstrates that a position-control-based robot could also achieve dynamically stable and robust locomotion. With the help of offline dynamic-model-based trajectory optimization algorithms and online simplified-model-based reactive controller, the tested quadruped robot Pupper achieved dynamically stable trot gait both on flat grounds and high slopes of at most 20 degrees. It also gained the ability to trot on and off a 10 mm plank blindly, proving the control framework's effectiveness as well as the potential of a position-control-based robot for indoor or structural environments.