> I have a document fragment that I need to parse with an XMLFilter.
I may be misinterpreting your question, but... Parsing's really Xerces' problem rather than ours... The easiest way I know to parse an XML Fragment right now is to make the parser see it as a complete document; write a stream wrapper which generates a dummy start-tag and end-tag around it, parse that, then remember in the rest of your processing that the dummy element isn't really part of what you're trying to process. (I believe there's an XML Fragments working group which is supposed to address how one represents a chunk of data extracted from a larger document, and I believe this is essentially the approach they're taking. It's a lot easier than inventing a new set of parsing conventions.)
