* James M Snell <[EMAIL PROTECTED]> [2005-08-04 21:15]: > Personally, however, I think that the elegance and simplicity > of in-reply-to and replies link rel values trumps defining them > as elements in a separate namespace or an otherwise perfectly > engineered solution. As for specifying the source of the > resource being replied to, a namespace extension inside the > link element is probably the best way to go.
I consider it stupid blunder to have proposed using atom:link for this, not elegant and simple. You need a link and an extension element either way. So choose to squirrel away information that even clients without threading support could use inside a namespaced element, and instead put information they have no use for in a place they can access. What a mess. Why not do it the other way around? > Using the via link just seems unnatural to this purpose in that > we're not expressing a relationship like "The information > contained here-in was received through..." we're expressig "The > information contained here-in is a response to...". Via works > great for the former and doesn't fit for the latter. So just drop back to “related.” Either way the exposed information is sensibly usable even by clients without threading support; they don’t know *why* it’s related, but it certainly is useful to let them partake of that information. > All of the examples that have been offered built around the via > mechanism just seem entirely too verbose for comfort... Verbose, sure. Too verbose, how? > especially given that we're trying to optimize for optional > metadata that hints at a relationship that can't be reliably > confirmed (e.g. that the entity being replied to MAY exist at > the specified source). You are going to need a namespace any way you turn it. If you really don’t care about exposing as much information as they can use to as many clients as possible, and brevity is such a concern, then maybe it should be <thr:in-reply-to src="http://example.com/atom">urn:foo:1</thr:in-reply-to> which makes this a Structured instead of a Simple Extension Element, but then maybe that was unavoidable anyway. Of course, conscientous producers will drop in a <link rel="related" href="http://example.com/atom" /> for good measure, so it’s verbose anyway. Hmm. Huh. Interesting. I like this even better than my previous proposition. D’oh. Screw sticking to Simple Extension Elements, this is better. This is *natural* – no funny business with markup inside atom:link elements. Each piece of information is available to as broad an audience as possible: threading-aware clients can correlate the “related” links to the source attributes, and non-aware clients can at least tell there is something related at the indicated resource. Regards, -- Aristotle Pagaltzis // <http://plasmasturm.org/>