W3C home > Mailing lists > Public > www-archive@w3.org > April 2009

Re: Review Comments for draft-nottingham-http-link-header-05

From: Sean B. Palmer <sean@miscoranda.com>
Date: Fri, 17 Apr 2009 12:29:35 +0100
Message-ID: <b6bb4d890904170429t3ab875abr853970e747a57b61@mail.gmail.com>
To: Julian Reschke <julian.reschke@gmx.de>
Cc: HTTP Working Group <ietf-http-wg@w3.org>, www-archive <www-archive@w3.org>
On Fri, Apr 17, 2009 at 11:34 AM, Julian Reschke wrote:

> The disadvantages are: [...]
>
> - incompatibility with RDF properties

If you want compatibility with RDF properties, the specification will
have to state that all URI extension relations MUST return a 303 See
Other response when dereferenced, as a consequence of this finding:

http://lists.w3.org/Archives/Public/www-tag/2005Jun/0039

Otherwise, you will have the situation where there can be valid URI
extension relations which aren't compatible with RDF, and that the
only way to test will be to dereference and place burden on the
server.

> and Atom link relations.

Atom link relations are IRIs, whereas the Link header extension
relations only cover URIs. So there's a similar incompatibility case
here, though at least this doesn't require a server dereference to
test.

Overall I see the case that you're making for URIs, but it would be
great if this could be tempered with some prose that mitigates against
the inherent disadvantage of using URIs which Anne van K. has pointed
out.

> It seems that clarifying the use of meta/@http-equiv is something
> the HTML WG should do.

The basic problem here is what happens when an HTML 2 to 4.01 document
is accessed on the filesystem. The behaviour of user agents peeking at
http-equiv is, as Anne van K. points out, not allowed by the HTML 4.01
specification (and none others back to HTML 2; I've checked). But it's
certainly the case that, for example, browsers will try to sniff the
encoding from an http-equiv value for Content-Type when the HTML file
is on the filesystem.

So this is a de facto thing. I'm not sure whether this user agent
behaviour is limited to sniffing an encoding or whether there are
other things that get deployed. Refresh, potentially, though again
this is something that HTML 4.01 explicitly tells people not to do and
yet is still, as far as I know, widely supported.

Kindest regards,

-- 
Sean B. Palmer, http://inamidst.com/sbp/
Received on Friday, 17 April 2009 11:30:16 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 7 November 2012 14:18:21 GMT