When computing a name 'from content' for a link, the algorithm walks the subtree and incorporates pseudo-element and CSS-generated text. Which detail of this traversal is correct?