An event correlation is becoming an important service in event-based middleware allowing subscribers in publish/subscribe paradigm to consume patterns of events (composite events). Recent evolution of wireless networks makes events flow from tiny sensor networks to Internet scale peer-to-peer systems among event broker grids. This new paradigm requires composition of events in heterogeneous network environments, where time synchronization and network conditions vary. Most extant approaches to define event correlation lacks a formal mechanism to define complex temporal relationships among correlated events. Here, we introduce generic composite events semantics introducing interval-based semantics for event detection supporting resource-constrained environments. We precisely define complex timing constraints among correlated event instances. We discuss underlying time systems and outline real-time temporal event ordering.