The proposed RM does not make particular assumptions about how knowledge is fed into an IMMPS. With regard to the expert components of the knowledge server, we just note that there are a several different ways to fill them with knowledge. As a portion of the knowledge stored in the expert modules is static in nature, it may be directly coded by the system builder, or only once read in from an external source (via the Acquisition Interface). However, it may also become necessary to acquire some knowledge during runtime, e.g., from the user. In an IMMPS such interactions can be handled in at least two different ways. One approach is to equip the Acquisition Interface of an expert module with an own user interface, so that the module can itself initiate a clarification dialogue. For example, there are systems which simply pop up form-sheet style windows, in case some knowledge is missing. While being simple, this method has the drawback that the clarification dialogue is not well integrated into the presentation, and thus can be rather disturbing. A more ambitious approach to the problem is to handle the acquisition of lacking knowledge as an additional presentation task. Basically the idea is to design an interactive presentation in such a way, that the user will provide the missing information when interacting with the presentation. For example, if the user is enabled to undertake walk-arounds in a 3D-graphics which is part of a multimedia presentation, the system may learn about the user's information needs and viewing preferences without explicitly asking her/him.