Event processing systems have wide applications ranging from monitoring RSS feeds to managing events from RFID readers, and there exists much work on them in the literature. Many competing temporal models for event systems have been proposed, with no consensus on which approach is best. In this paper we determine the important properties for such temporal models. Our approach is to define a very general temporal model capable of representing time in all of the major event systems. We introduce axioms motivated by the time stamp ordering relation and the semantics of the successor operator, which is present in all event systems. Only two of our axioms are controversial; the remaining axioms are satisfied by all event systems. We consider the temporal models obtained using our full set of axioms, and the models that result when one or the other of our controversial axioms is weakened. In one case we see that there is no acceptable temporal model. In the other two cases, we show that the resulting temporal model is effectively unique up to isomorphism, leaving us with only two different models. Finally, we argue that one of the two models is better than the other when both naturalness of semantics and efficiency of implementation are considered.