The real problem, and I use this article and your post as evidence, is that after more than a decade, XML is poorly documented, poorly understood and hard to use. The tools stink, and unspecified proscriptions to "use good tools" pervade the XML community yet nobody seems to really be writing "good tools" and there seems to be very little consensus about what is a "good tool".

What I mean by poorly documented is this, to people unfamiliar with XML, they head over to their local bookstore to pick up a guide or somesuch and find a 2000 page tome with 20 authors on the cover setting forward a several month path of reading and writing code to understand the concepts in the book, and probably finding little illumination on the topic to help them grok the part of the vast XML spec they need to parse and map relatively simple data structures to their software -- something that should take a half page tutorial in some random XML library's documentation.

And I agree, DTDs and XSD are in a sense external to XML, but when every 2000 page tome you pick up hammers away about XML validity, you feel like you have to support those too -- and supporting those adds vast amounts of complexity to your application, the XML format you are using, the message interchange, etc. (so nobody who knows anything uses them anyways).

So what should be a 2 hour development task turns into a 3 month hair pulling and library testing exercise.