Pushing DTD over XML Schema is pretty surprising. A DTD can't express namespaces correctly (documents aren't required to use the prefixes you hoped they would) nor constraints like "this attribute must contain a timestamp".

It should also be noted that the JSON Schema situation is pretty grim, with an expired Internet-Draft that has seen almost no live use, and that presents all the same problems as using XML with no schema (only worse, because there's also no standard for avoiding name collisions).

I don't think the author was particularly advocating DTDs over other schema systems. The real point is that schemas make XML workable with other tools.

Exactly that was my point.

On a tech topic worth talking about, in the difficult mode of "laying down the law", that's some pretty fine writing that says something that bears repeating and refinement.