Disk-based XPath Processing

Is there something like this? If so, how is it implemented? Is it  
even reasonable?

To clarify: I'm looking for approaches - preferably available in the  
form of an open source library - that allow for running XPath  
expressions against an XML document without the memory requirements  
of creating a full tree in RAM.

Any pointers would be greatly appreciated.

Stefan Tilkov
--
Stefan Tilkov, http://www.innoq.com/blog/st/