Comments received on the Self-Describing Web Finding

I've received some email comments on the latest draft of the Self-Describing Web finding. This note gives pointers to some of the comments that have been received, primarily so we'll have a handy list for reference during the Feb. 2008 TAG F2F.

Emails with comments on Self-Describing Web

Some of the emails that have been received commenting on this draft include:

Dan Connolly, 19 Feb 2008, http://lists.w3.org/Archives/Public/www-tag/2008Feb/0077.html
David Orchard 26 February 2007 http://lists.w3.org/Archives/Public/www-tag/2008Feb/0094.html
Mark Baker 27 February 2007 http://lists.w3.org/Archives/Public/www-tag/2008Feb/0096.html

Highlights of comments

Some of the most significant comments appeared to be

(Dan) "Good Practice: Web resource representations SHOULD be self-describing." That doesn't stand well on its own. I was prepared to be a bit generous and let "self-describing" stand for the several paragraphs around it, but then I got to the 2nd GPN: "Good Practice: Use the HTTP protocol to deploy self-describing resources." If that's not redundant w.r.t. the 1st GPN, I don't know how to understand the 1st GPN.
(Dan) Please cite issue 51 from "Similarly, the need may arise to use new values for individual fields such as rel attributes on HTML link elements."
(Dan) Please cite issue 51 from "Similarly, the need may arise to use new values for individual fields such as rel attributes on HTML link elements."
(Dave, also endorsed by Mark) I still have problems with the RDF uber-alles statement of "RDF [RDF] plays an important and distinguished role as the preferred technology for creating self-describing Web data resources". I disagree with the "preferred" part. Ironic that this assertion came right after you showed how Atom was supporting self-describing Web data resources".
(Mark) I'd suggest not using "application/xml" in your example in sec 2 (or in 4.4) because in practice it's used primarily in a non-descriptive way, i.e. as a catch-all for all XML content with a (non-grounded) presumption that the transfer semantics are indicated by the namespace of the root element of the document, rather than by using a more specific XML media type. You may even want to mention this as an example of how not to be self-descriptive, and include a reference to the authoritative metadata finding.
(Mark) in section 4.1, it might be getting too detailed, but perhaps you could mention the relative cost of deploying a new element from each layer, justifying a best practice such as "if you can define your extension as a new media type, prefer that over defining a new URI scheme", etc..
(Tim - private communication - paraphrased) Why is RDFa positioned as a good example of self-describing content? It's not an official standard and not widely accepted by the community. It's not clear you can really follow your nose and have it work at this point.