XML is markup

Text plus tags There's nothing magical about XML. It uses the same notion of tags intermingled with text that was used in SGML, HTML, and a number of other languages. The text provides the content, the tags provide the names and structures that apply to that content.
Labels as types XML 1.0 has very little built-in notion of types, though it supports a few pieces (like ID, IDREF, and IDREFS) which provide additional constraints on content. Instead of saying "this is an integer", XML might say "this is a population" and the program would have to know it was an integer, if it in fact cared.
Not enough for the 'dataheads' While XML addresses most of the concerns of the document-using and managing community quite well (with the exception of support for overlapping markup), people used to working with databases or objects just didn't see how this text and structure stuff could possibly be heavy-duty enough for their applications.