Introductory questions: ref counting, libxml access

I am evaluating Gdome, and I had a few introductory questions:

1. What gets freed with gdome_di_freeDoc? Basically, I would like to
avoid the explicit reference counting if possible: parse a document, use
it all I want, and then free all the memory at once when the document is
no longer being used (sort of the libxml model).  The examples seemed to
be mixed in their use of the reference counting mechanism, and I decided
to try the mailing list before reverting to valgrind to find out what is
really going on.

2. The DOM implementation seems refreshingly complete, but many
operations in other libraries require libxml structures (for example,
XSLT and XPath). I would like to be able to use a DOM document and then
transform it without reverting to non-portable hacks like copying the
gdome-xml-*.h data structures. Is there a method to get the underlying
xmlNodes that I am missing?

3. Is there a method to convert from an xmlDoc to a DOM document, and
xmlNode to a DOM element, etc.?

Thanks,
Bret Taylor
[email protected]