Validating streaming XML documents

Abstract

This paper investigates the on-line validation of streaming XML documents with respect to a DTD, under memory constraints. We first consider validation using constant memory, formalized by a finite-state automaton (FSA). We examine two flavors of the problem, depending on whether or not the XML document is assumed to be well-formed. The main results of the paper provide conditions on the DTDs under which validation of either flavor can be done using an FSA. For DTDs that cannot be validated by an FSA, we investigate two alternatives. The first relaxes the constant memory requirement by allowing a stack bounded in the depth of the XML document, while maintaining the deterministic, one-pass requirement. The second approach consists in refining the DTD to provide additional information that allows validation by an FSA.