We describe a tree difference and edit sequence algorithm with expecte
d sequential execution time O(nlogIogn) and expected parallel executio
n time O(logn), for trees of size n. The algorithm assumes unique labe
ls and permits operations only on leaves and frontier subtrees. Despit
e this assumption, its application domain includes structured text and
hypertext, and it is much faster than existing algorithms based on dy
namic programming.