The Kalman filter as posterior Bayesian reasoning