The HTML parser creates two nodes with tag=html

[burg]

One of them seems to correspond to nothing, and has the Doctype node and HTMLHtmlElement as children. The other represents the HTMLHtmlElement itself.

Unless there's some way to query the doctype node through the tree structure from script, we should save doctype somewhere and submit the HTMLHtmlElement node to layout as the document root.

[bzbarsky]

Script can most certainly get at the http://dom.spec.whatwg.org/#interface-documenttype

[bzbarsky]

But for purposes of layout, of course, the doctype is irrelevant. The "nothing" node you have there is also reachable from script. It's http://dom.spec.whatwg.org/#interface-document

[jdm]

This is fixed now.