SMS: On Markov policies in continuous time discounted dynamic programming /