Educational institutions face the recurring problem at the beginning of every semester to determine when classes and examinations take place. This is not an easy task as dozen of constraints and wishes need to be taken into account. So far the literature on timetabling algorithm design has almost exclusively focused on the short term: given input on requirements and wishes, how can we make the best possible timetable for the next coming semester? There are two concerns with this practice. Firstly, the performance of algorithms depends on the requirements and wishes that need to be taken into account. Hence, an important question is what algorithm a practitioner should use. Answering this question is complicated by the fact that the algorithm should work well for a number of future semesters for which the specifics of the timetables to be constructed are not yet known. Without a long-term perspective on algorithm performance, the choice of algorithm a blind guess, and its results could be poor and costly. A second issue relates to the common (but false) assumption that timetables from different planning periods can be generated independently from one another. A concern with this practice is that there is no guarantee about a fair distribution of timetable inconveniences over stakeholders and time. Hence, we propose the construction of timetables for multiple semesters at once, thereby taking into account that the complete problem input for future semesters is unknown.