AUTHOR = "Jeffrey Palm and Karl J. Lieberherr",
TITLE = "Improving XPath Evaluation with Strategies",
INSTITUTION = "Northeastern University",
YEAR = 2005,
MONTH = "September",

Paper .

Notes by Karl:

Can theorem 1 be strengthened from exponential (2**n) to faster growing functions?
For example, can we construct an infinite sequence of query/DTD/document triples
so that the naive evaluation will visit A(c,n) nodes, where A is the Ackermann function
and c a constant > 3?