Danny Ayers makes the link between blogging and annotation. I've been thinking about this sort of thing from a different (although related) viewpoint.
Lately I've been thinking about implementing comments and/or trackback in Leonardo. I personally think they are essentially the same thing and that there is a obvious relationship with web page annotation.
The way trackback works is you ping a blog with information about a reference that has been made about an entry in that blog. The information can include an excerpt of what was said.
The trackback implementations I've seen tend to give a trackback URI for an entry that is different from the URI for the entry itself. A more RESTful approach (and one I plan to implement in Leonardo) is to have the trackback URI be the URI of the entry. So you POST to the blog entry to trackback.
I had already considered POSTing to the blog entry as the mechanism for comments and that is when it first struck me that comments and trackbacks are really the same thing. The fields that you POST would be slightly different, but the mechanism should be the same.
Which leads me to web page (or, more generally, resource) annotation. There is no reason why the resource you post to should be restricted to being a blog entry. In fact, in Leonardo, there is nothing special about a blog entry—implementing trackback/comments for blog entries would enable the same capability for any page on my site.
Finally, in all these cases the mechanism involves POSTing the comment/trackback/annotation to the source itself but there is no reason why such information couldn't also be detached and expressed in RDF for the purposes of annotation servers, Technorati-style sites, etc.
I'd like to see a spec that supports this approach. I don't think it should be Atom's initial goal but I would like Atom to at least be compatible with the kind of unification I'm describing.
UPDATE (2004/05/06): see Joe Gregorio's response.