Caching source documents and key()s

Hello All,
this is not exactly about the XSL language...

I have a JAVA web application which has to present content from
the same XML document several times for an user task, sort of
a poor, short-lived in-memory database.
Can anybody share experiences on caching source documents,
using TrAX/JAXP1.1? Should i cache the File representation,
or is it better to cache a DOM tree or SAX events, provided
that ca. 60% of the file content is markup or ignorable whitespace?

Furthermore, i use key() heavily to pull stuff from cross referenced
elements. It would be nice to keep already built indexes across
invocations of the processor, preferably along with the processor
dependent representation of the input source. Something like what
Templates is for the style sheet.
Any comments from our processor implementors about this?

Regards
J.Pietschmann

 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list