Well, in a way, he is right. The XML specification does not address separation of content from format, or most other design decisions, and XSL-FO is in XML document syntax and is about page formatting. Yes, but.
XML was written as the ‘good parts version’ of SGML. It is a short document that concentrates on document grammar and well-formedness; DTD grammar and syntax, validity constraints, entities, processors, and the like. It assumes that you want to know how to define a markup vocabulary and says nothing about:
Why tag information
Separating formatting/look-and-feel/behavior from content
Procedural markup versus descriptive markup
Generic markup
Structural markup
Semantics
Best (or even recommended) practice for modeling vocabularies
Data versus documents
Architectural forms and reuse
Separation of concerns
What should or should not be expressed in a grammar
Element versus attributes
Document-based XML is assumed, by large numbers of its practitioners, to include:
Generic Markup To many of us, the idea of XML incorporates “Generic Markup” that is, the art of identifying information by what it is, not by what it should look like at any moment or how it will be used in any one place. When making up tags for “regular documents” (what ever that means), it is generally considered better if you use generic markup rather than format-based markup or behavioral, procedural markup. Much, but not all, XML benefits from an underlying skeleton of generic markup.
Separating format from content The principle of separating content from format (what it is not what it looks like) is another formulation of the same general idea, although it goes a bit further in that it implies that there can be more than one display or use for the same content. This is, in fact, one of the main virtues of XML as it is generally used. We can take the same tagged text, extract from it, superset it, repeat anything we need, delete what we want, and each of us make it look-feel-behave in an entirely different way.
Separation of concerns Some people think there should be a schema with a widely used grammar for a document type, with perhaps several sets of Schematron rules to enforce more local business rules that are also imposed on those documents, other validations as necessary, and then we can each use a different stylesheet to make it look like we want. There are long discussions of what should be imposed based on a grammar, and what should not.
But the XML Recommendation mentions none of that.