Complex event processing aggregates and combines information from sources into higher-level information or knowledge at the appropriate point through analyzing relationships among instances of various event types. Although composite events have been a useful modelling tool in active database research and network monitoring, little progress has been made in complex event porcessing middleware. The time model including temporal ordering of events is a crucial aspect for complex event processing. The semantics of operators for complex events is not defined in a uniform manner in existing middleware and applications, leading to the inconsistent results from the desired semantics and the detected complex event instances, and there is need to trade off time model against event detection efficiency. To address these issues, a consistent time model is defined in this paper for complex event in the service-oriented computing platform InforSIB. It includes solutions for time model of complex event, unsynchronized timestamp and out-of-order event arrivals. Basic operators provide the potential of expressing the required semantics and are capable of restricting expressions by parameters. Interval-based semantics for event detection is introduced, and extended, defining precisely complex timing constraints among complex event instances. The corresponding efficient complex event detection algorithm is presented too. And experiment results prove its effectivity.