As already pointed out by Ruurd, there's an excellent article on Groklaw that compares the OpenDocument Format (ODF) with MS-XML (Microsoft's new "XML" format for MS Office documents (note the "")). At work, I have architected, esigned and implemented (with the help of two coworkers) an application that generates a reference guide from the configuration data of our platform. It generates both XHTML and OpenOffice XML (will most probably be ported to ODF soon, but the format is almost identical anyway) with pluggable backends, which means I've been digging into the OOo XML format since quite some time and it's a nice format to work with. Not perfect from a parsing point of view, but nice nevertheless (e.g. DocBook XML would have been even better, as it's hierarchical for section elements whereas OOo is not):
<section title="one"> <section title="two"> <para>...</para> </section> </section>
<text:h text:style-name="Heading 1" text:level="1">One</text:h> <text:h text:style-name="Heading 2" text:level="2">Two</text:h> <text:p text:style-name="Text body">...</text:p>Still magnitudes better than MS-XML. For those who had the dubious luck of looking at the HTML that is generated by MS Frontpage, you might see the same parallel as I do with MS-XML. Non-mixed content is awful and not even near (X)HTML. Well, what would we have to expect from Microsoft anyway...