Distributed queries without distributed state.

Traditionally, distributed queries have been optimized centrally and executed synchronously. We outline a framework that relaxes both of these constraints using mutant query plans: XML representations of query plans that can also include verbatim XML data, references to resource locations (URLs), or abstract resource names (URNs). Servers work using local, possibly incomplete knowledge, partially evaluate as much of the query plan as they can, incorporate the partial results into a new, mutated query plan and transfer it to some other server that can continue processing. We present preliminary performance results, and discuss issues and strategies for mutant query optimization.