XPath [3, 5] is a powerful and quite successful language able to perform complex node selection in trees through compact specifications. As such, it plays a growing role in many areas ranging from schema specifications, designation and transformation languages to XML query languages. Moreover, researchers have proposed elegant and tractable formal semantics [8, 9, 10, 14], fostering various works on mathematical properties and theoretical tools [10, 13, 12, 14]. We propose here a novel way to consider XPath, not only for selecting nodes, but also for tree rewriting using rules. In the rule semantics we explore, XPath expressions (noted p, p ) are used both on the left and on the right side (i.e. rules have the form p → p ). We believe that this proposal opens new perspectives toward building highly concise XML transformation languages on widely accepted basis. Categories and Subject Descriptors D.3.1 [Programming Languages]: Formal Definitions and Theory—document transformation...