Richard Saul Wurman used to say "Clarify, Don't Simplify". This article proves that point. It is simplified to the point of distortion.
1) XML is nothing in and of itself. It is a metalanguage; i.e., if you want to make a markup language, if you do it this way, then it is XML.
2) Everything depends on your instantiation of XML. In other words, the DTD or Schema that is used is much more critical to the success of your efforts than the various tools mentioned in the article. Designing the DTD/Schema (i.e., the data model) is almost never mentioned, yet that is so much harder to do right than anything else.
3) XML is a lousy data store. It biases towards a hierarchical data structure, which went out of style in the 1980's for a reason. It's incredibly bloated (I remember file sizes 43x the size of the same data stored in a flat file), and the pre-built infrastructure pales in comparison to relational databases.
4) What infrastructure is there is very immature, even after all these years. I was working on the FDA's SPL initiative, which was based on an XML Schema. XML Spy compiled the schema as an XML file, while another tool found errors in the schema as a schema. What good is that?