Recommendation please: forming an XML toolkit

[jmdeschamps at gmail.com]

I'm asking help on this topic of building a tools foundation for future
XML projects because of the sheer volume of possibilities I will not
have the time to check/try/test in detail.

This question is a bit like the ones pertaining to 'Which web framework
to use?', there is a lot of good stuff out there, and often it boils
down to personnal preference, mind-fitting interface and such BUT... to
make it more precise I will give more context on the future projects
involved...

I've done some homework trying out a few packages based on published
tutorial: Boddie's Python and XML: An Introduction (for minidom),
Lundh's elementTree examples, I read a bit about Amara, pyXML, others.
I've read a bit on ease-of-use, benchmarks, pythonesque versus
job-protection  perspectives,
I've even tried building my own xml2PythonObjects2xml tools
Finally, I've read-up a few threads pertaining to the question 'which
XML packages to use' !

Some considerations I have using XML :
1- representing inter-connected academic articles in text-based files
whitout a sopecific BD package
2- being 'easily' able to modify the structure of these documents as
search tools evolve
3- searching through these articles often, and with evolving
algortihmic complexity (from word base search to RDF-type meta data, to
OWL-type semantic information, etc)
4- In a context where I'm (in a practical form) evangelizing the use of
Python as a great tool from going from 'this could be a new approach'
to 'this piece of code realizes that approach'
 (managers need to be confident that they could choose to make this set
of articles evolve using other languages (other developpers not caring
for Python for example, but Java instead)  and the infobase would be
directly accessible and algorithms understandable from a 'popular'
xml-manipulation point-of-view (Using DOM, I guess a regular DOM-SAX
Java developper would understand Python code, but would they if the
code relied heavily on elementTree (for example)
5- relying as less as possible on complex third-party libs (to use A,
first get B from elsewhere, which itself requires C from still another
place...) I DON'T mind the simple package (PIL comes to mind here)
6- VERY IMPORTANT CONSIDERATION - That I can keep my focus on
developing algorithms, MINIMIZING XML clutter I don't want to become a
XML guru - I like the simple principle of XML as a tree of tagged
elements that have attributes and text data...

Thanks for any and all who read this, and those who have experience
ressembling what I'm about to embark on for your help !

Jean-Marc