It's not as dramatic as the article makes it seem.
XML is and was a great solution far a huge number of things. Overexcited developers were tad too eager to put it to use for things outside its true scope (regardless of what it was marketed as), and now they're realizing that there are better alternatives for those particular applications.
XML was and still remains an excellent solution for the problem it originally solved: a joint human and machine readable markup language.
Uh... a tad too eager? Dude, we crossed that threshold with Ant or JSP. By the time XSLT and XQuery rolled around, we were looking at a full-on stampede of developer group-think.
XSLT happens to be my pet peeve. I simply can't understand how anyone would have ever looked at that problem ("how to turn a source data document into a presentation format" -- something that has been solved sanely a thousand times by obscure technologies like "scripts", or "PHP") and decided that the best way to handle it was a turing-complete pattern matching language written in XML itself! I mean, it looks more like a torture device than a programming language...
Yes, XSLT written in XML itself is abominable. It's easier to read if the program and data have different formats (lisp notwithstanding...). But it's escaping less-than signs that got really me: (i < 5).
But transforming trees with a functional programming approach (not PHP) is a natural fit.
XML is and was a great solution far a huge number of things. Overexcited developers were tad too eager to put it to use for things outside its true scope (regardless of what it was marketed as), and now they're realizing that there are better alternatives for those particular applications.
XML was and still remains an excellent solution for the problem it originally solved: a joint human and machine readable markup language.