Considering the Wireless Sensor Network's classical necessity of sensor-to-sink communication, data aggregation can be applied to combine data coming from different sources enroute to a simple sink, allowing in-network consolidation of redundant data. Moreover, it has been put forward as an essential paradigm for WSNs to: conserve the scarce energy resources, minimize the number of transmissions, and eliminate/reduce redundancy. Our objective is to identify the specific characteristics and the potentiality of data aggregation in wireless sensor networks and discuss how existent approaches have implemented it. In particular, the work proposes a general taxonomy of data aggregation mechanisms, while discusses their design alternatives. In this way, we hope to provide insights that will help the choice of which mechanism is more suitable for a particular task. We notice that mobile environments have not always been considered in the discussed data aggregation approaches, and so represent an interesting open issue.