give back untreated XML if possible: add XML header, refrain adding sugar
What about a simple possibility to get back just the untreated content (the parsed markup) of a page, if necessary just prefixed with a simple <?xml version="1.0" encoding="UTF-8"?> header?
This means: Let the parser decide what is given to the user; do not add any wiki sugar to the page.
Hm: A simple fix could be to change the header of the print action to xhtml, even you cannot guarantee that the parser's result actually ist xhtml (or even well formed). (I can't provide a patch, because the print action seems to be narrowly knitted into the request class.)
I started some attempts here proposing to make moin more XML compatible (see XmlWiki) . Well, I understand now that this is big stuff, though I do not agree that this issue should have low priority 
However, since there might be other users like me who want to read out wiki pages with xsl or other XML compatible tools, this fast compromise would sweeten the waiting.
What I need the XML header for: I'd like to read out a page which is created from a group of peoble building a thesaurus (see http://www.ews.uni-heidelberg.de/~busse/moin.cgi/SchlagwortNormDatei?action=raw). An XSLT-processor produces a rich index then (c.f. http://www.ews.uni-heidelberg.de/~busse/moin.cgi/Sw). PROBLEM: The applied pseudoXML parser gives back well formed XML, but action=print adds html headers (and a lot of more ill formed html tags (e.g. meta)). This makes postprocessing more complex.
Thanks! jbusse
MoinMoin already supports XML output. Have you checked out the XML formatters that are integrated? See HelpOnActions on how to call them.