Communication latency plays a key role in time critical control applications in dynamical cyber-physical systems. It is in particular influenced by source coding algorithms which are typically developed independently of the process dynamics driven at the nodes of networked control systems. This is a note on model-based source coding, providing a state-dependent coding strategy. Therfore we confine our discussion to the simple case-study of Markov chains. Our goal is to minimize the expected codeword length of transmitted information by accounting for the process dynamics.