Schedulers and finishers: On generating the behaviours of an event structure

It is well known that every trace of a transition system can be generated using a scheduler. However,this basic completeness result does not hold in event structure models. The reason for this failure is that,according to its standard definition,a scheduler chooses which action to schedule and,at the same time,finishes the one scheduled last. Thus,scheduled events will never be able to overlap. We propose to separate scheduling from finishing and introduce the dual notion of finishers which,together with schedulers,are enough to regain completeness back. We then investigate all possible interactions between schedulers and finishers,concluding that simple alternating interactions are enough to express complex ones. Finally,we show how finishers can be used to filter behaviours to the extent to which they capture intrinsic system characteristics.