We describe the design and preliminary analysis of an optimization technique for XPath called warp-edge optimization. The XPath data model is a tree-like data model that has an edge from an element to each component in the content of that element. The edges are traversed in the evaluation of an XPath expression. A warp edge is an edge that is something other than a parent to child edge, i.e., an edge from an element to a sibling or to a grandchild. Warp edges can be dynamically generated and stored during query evaluation to improve the efficiency of future queries. We describe the implementation of warp- edge optimization as a layer on top of Xalan, the XPath evaluation engine from Apache. Experiments demonstrate that in the evaluation of some XPath expressions, the use of warp edges results in substantial savings of time.