Sam also asks if he should have TrackBack RDF in his main page. I say no, for two reasons.
First, it just encourages foolish behavior: if you are pinging a TrackBack entry you are replying to a specific post, and so you should be linking to that post, but if you click your bookmarklet on the main page you automatically get the link to the main page, and then have to go back, copy the permalink for the post, back to the bookmarklet, select the URL in the automatically inserted link, and paste over it. Especially in a case like Sam’s, where reading the comments also means going to an appropriate URL to link to (I just noticed that it isn’t actually the same URL as the permalink), the only reason for pinging from the main page is to say “I want your readers to read me, but I can’t be bothered to send my readers to your specific post, and what’s more I can’t be bothered to even see what other people said about your post.”
As if that wasn’t enough, at least in MovableType’s case, not having the TrackBack data in the main page would allow pointing to the RDF in a separate file with an HTML <link> tag, rather than embedding it (btw, the RDFCore WG no longer recommends embedding RDF in HTML), which neatly solves the problem of validating a mix of RDF and HTML. MT can’t do <link> tags for the main page, because they have to go in the <head> section of the page, but MT doesn’t actually know what posts will appear in an index page until it generates it, while any archive type that can be used as the target of a permalink does know what posts it will include before it generates them.
Apparently during the TrackBack beta, people were puzzled by why they couldn’t ping from a main page, so Ben caved. I don’t see why, and if I wasn’t using the RDF in my main page as an example of my methods for fooling the validator, I’d take it out.