> link text (inline textual content). Generalizations break the idea instead 
> of making things more flexible. What will a list of links look like if a 
> "link" can be any element, including the root element? Similar 
> considerations apply to embedding.

There is, however, market demand for such features, whereas there is very
little demand for the ability to extract lists of links (its probably seen
as undesirable, as it allows you to bypass all the clever page design).