An update operation presupposes that one can predict how the world changes along time. In the absence of a predictive model of evolution, the common sense law of inertia is usually used and justifies the minimal change approach to the frame problem. Instead of relying on an implicit modeling of persistence, we propose to use an explicit modeling of the expected evolution. We first explain why our previous proposal was not quite satisfactory. We then propose a model-driven semantics of the update operation which can take into account an explicit transition model. The transition model provides a more powerful and flexible way to represent the persistence of information. It can moreover be stressed that non only persistence information but also default and transition rules can be expressed in this transition model. It can then be used to represent evolutive systems as required for example in a monitoring or diagnosis context