Peter,
XLink was and always has been a partial solution at best, and at worst it has been extraordinarily counterproductive. XBRL to me is a very good indication of what happens when XLink is used with abandon - hideously complex documents that were expressed in XML but were so heavily linked that even an XML database struggles with the joins, split across five or six different documents, and relying typically upon relational tools to generate. It would have been a good candidate for RDF except for the fact that RDF was mostly an academic exercise when the spec was produced, and now the language is too heavily entrenched to change.
I actually have a suspicion that RDF may end up enjoying a renaissance on the web, and that in turn may very well put some pressure back into the question of linking. If I create a construct like
<p>All content on this site is licensed under
a Creative Commons License</a>. ©2011 Alice Birpemswick.</p>
then this provides a declarative hook for linking with semantics attached, which really was one of the key benefits of linking in the first place.
Kurt