As data stream springs up in various areas, distributed data stream management systems are being paid more and more attention. In DSMS, join is one of the most common but complicated operators, and the efficiency of continuous queries may be influenced by join directly. This paper mainly studies the join operation over data stream located on two different sites in WAN. Firstly, data model, query model and cost model are defined, and then a simple direct-join strategy and a semi-join strategy are proposed, which are respectively ideal under certain conditions but can't adapt to the variation of data stream. Finally, based on the analysis of the two strategies, an optimized adaptive join strategy is presented. The efficiency and flexibility of our strategy is proved by extensive experiments. Key words: distributed data stream management, continuous query, adaptive join.