Links on the Web aren't standard. There are standard basic cross-document links and some of what I consider hacky intradocument linking elements, but linking as a whole isn't yet standardized.

<a href="mypage.html">text</a> is a basic cross-document link. The rel and rev attributes may further detail the link.

<link href="mypage.html"> is sort of the same thing as the basic a href link, in terms of semantics, but instead of text being the entity that relates to the linked document, the entire document relates to the linked document when a link element is used.

dl>dt and dl>dd are links. The contents of the dt elements are linked to the contents of the dd elements. Unfortunately, this format places considerable limitations on the uses of these links, because they must be in a list with no content in between.

h1 through h6 are links. The content of the hn element links to everything that follows until the next hn element of the same n, or apparently in some cases (although the W3C fails to detail on this) until the hn's parent element closes.

There are also many other types of links, such as created by the label, title, captain, tr, etc. elements. You might even consider the elements abbr, dfn, q, etc. links, although they are extremely restricted.

Believe it or not, it's actually possible to combine all of the above types of links into one element and a small set of link classes. Since the W3C's term link type is vaque and can be misleading, I will refer to those as link classes instead. Link classes may be used in link relationship descriptions such as the rel and rev attributes in some HTML elements. Link classes specified in HTML 4 include next, contents, help, and stylesheet, among others. This list would need to be enlarged slightly (and ideally modified in some areas) to accommodate for all of the different linking elements that exist in HTML. For example, there might be a definition link class or a source or origin link class, or a general use title link class that could more or less replace the title, h1 through h6, label, and caption elements by itself, since most of those really describe the same thing but for different elements.

Of course, a great deal of thought would be needed to come up with this list, and there would need to be a lot of discussion. As it is, even though the W3C is granting themselves complete freedom with a brand new backwards-incompatible version of HTML (XHTML 2.0), they are still planning on repeating several of the same mistakes they made in the previous versions of HTML. Most of these different limited linking methods remain, and they don't have a more powerful and flexible general use alternative that is native to the XHTML 2.0 core (XML allows embedding other XML formats directly into the document, so something like RDF/OWL could be used, but native methods would have more user agent support and would thus be better).

If the W3C truly plans to help build the semantic web, they had better think twice before choosing to forego this unique opportunity to standardize and enhance the methods of document and element linking, which is without a doubt the most important thing in building the semantic web.