According to the analysis [1], the material issues that SGML has with HTML are - lexical issues; in particular, id/href values can be almost anything in HTML

According to the analysis [1], the material issues that SGML has with HTML are - lexical issues; in particular, id/href values can be almost anything in HTML5, and are case-sensitive unlike other name-like tokens; note that while id/href values can be anything, they are however effectively constrained by other specification rules (eg. id values can't effectively contain quote characters because of attribute parsing rules) - script parsing is especially foobar'd (and injection-prone) in HTML (cruft from HTML3/HTML4 times) - HTML 5.1 constrains attributes on elements depending on the value of other attributes (possibly on parent element attributes), and also makes content model features dependent on parent attributes in some cases, which can't be described by an SGML DTD grammar (but then doesn't have to be described for parsing) YMMV but none of these things really matter that much in practice. [1]: http://sgmljs.net/docs/html5.html http://sgmljs.net/docs/html5.html