We propose a new algorithm to solve the problem of joining two spatial relations $R$ and $S$ when $S$ is indexed by an R-tree. The method combines classical query processing techniques and a plane-sweeping algorithm. We compare our approach to the simple indexed-nested-loop method and to the state-of-the-art algorithms, and show through analysis and experiments on synthetic and real-life datasets that our method outperforms in most cases the other ones, both in CPU and I/O. More importantly it appears to be more robust regarding the many parameters involved in spatial query processing.