W3C home > Mailing lists > Public > public-html@w3.org > March 2010

Re: New Version Notification - draft-nottingham-http-link-header-08.txt

From: Roy T. Fielding <fielding@gbiv.com>
Date: Tue, 2 Mar 2010 13:24:22 -0800
Cc: "Mark Nottingham" <mnot@mnot.net>, "Larry Masinter" <LMM@acm.org>, <public-html@w3.org>
Message-Id: <EF7A2AFF-2F08-403C-A997-CF4CDA072C8D@gbiv.com>
To: Krzysztof Maczyński <1981km@gmail.com>
On Mar 2, 2010, at 5:34 AM, Krzysztof Maczyński wrote:

>>> [1] http://lists.w3.org/Archives/Public/ietf-http-wg/2010JanMar/0168.html
>> 
>> 
>> But that's the whole point.  Your specific example of
>> 
>>> We are an author of resource A and we want to have a link between resources A and B. (We cannot or don't want to alter B or headers served with it.) To say that B is A's next, use rel and next. To say that B is A's prev, use rel and prev. To say A is B's next, use rev and next. To say A is B's prev, use rev and prev. The latter two (whose characteristic is that B is the starting resource) would be unavailable with the other interpretation of rev.
>> 
>> is inherently false -- one cannot create a link from B to A
>> on the Web with a Link header field on A.  The Web does not allow
>> such a feature to exist because it will not scale and is not
>> supported by the natural authority-relationship of DNS delegation.
> I'm… confused. (Especially having also read [1] and related TimBL's writings.) Of course a link expressed not at its source isn't directly suitable for traversal, so it's usually not a hyperlink (but those don't dominate among the use cases for this I-D anyway). (Sometimes it could be automatically turned into one. E.g. a browser could understand that if A is B's prev then B is A's next and navigate to B when a user requests the "go to next" function in the chrome. Or it could detect that B is rendered in another browsing context and on recognizing the link from B to A expressed in or along A (or even some C) update that rendering accordingly. I don't know, maybe HyTime, which informed XLink, has such features. That's something currently missing from the Web and I agree that the demand for it seems low (but see [2] (I vaguely recall seeing it in some requirements of planned features document; now I can only find this draft)).) But are you suggesting that decoupling expression of links (of any type, not just hyperlinks) from their participating resources and their roles is contrary to Web architecture? What about XLink, linkbases (whether using XLink or something else), RDF stuff (thanks for mentioning it - a triple may indeed be viewed as a link), SMIL [3]?

It isn't "contrary to Web architecture" -- it simply isn't supported by
the HTTP and HTML parts of follow-your-nose Web architecture.  A person
can have access to alternative sources of links (link bases, RDF graphs,
etc.) and make use of that information independently of the links
expressed via HTTP.  My point was that the Link header field does not
support those express third-party links because there is no way to bind
the relationship on the third party.  That does not prevent anyone from
defining the same links elsewhere, where hopefully there is enough
time and space to deal with the cross-authority issues or the user
has configured their view of the world/graph to overlay such links.

....Roy

> 
> Larry, has the TAG ever discussed this?
> 
> Best regards,
> 
> Krzysztof Maczyński
> 
> [1] http://www.w3.org/DesignIssues/Topology
> [2] http://www.w3.org/TR/2004/WD-SVG12-20041027/extendedlinks
> [3] http://www.w3.org/TR/SMIL3/smil-extended-linking#SMILLinking-ObjectLinking
Received on Tuesday, 2 March 2010 21:24:53 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 29 October 2015 10:15:59 UTC