POWDER Use Case (was Re: Reviving HTTP Header Linking: Some code and use-cases)

This mail is related to the thread started by Harry Halpin [1].

Acting on behalf of the TAG, Jonathan Rees has noted [2] that the POWDER
Working Group would very much like to see Link reinstated. Our relevant
Rec Track document [3] is due to be updated this Friday (tomorrow!) and
although we're changing the current text which more or less assumes that
Link will be reinstated, for an editorial note that says we'd like it to
be - HTTP Link is very important to POWDER.

Why?

In some respects, the Protocol for Web Description Resources is an
update of PICS [4]. Indeed, if and when POWDER is at Rec, PICS is likely
to be withdrawn. We expect agents to collect descriptions of resources
by querying repositories of such data but we'd also like to be able to
do a HEAD request on a given URI and see from the HTTP headers whether
there was a Description we could GET.

As for URI values for rel and/or profile... maybe things get a little
more muddy here.

Supposing you want to find out whether http://example.org/will render on
a mobile device. You do a HEAD request and get

Link: <powder.xml>; /="/"; rel="powder" type="application/xml";

So you follow the link, find the POWDER document and process it 'looking
for' a mobileOK trustmark (which will be encoded in this fashion). So
content providers may wish not only to link to a POWDER document but
also indicate which descriptive vocabularies are used in those
descriptions. In this case, the rel value might be "powder mobileOK".
And maybe they want to include ICRA descriptors for child protection
purposes as well so we'd have rel="powder mobileOK ICRA" and so on.

Replacing those words with URIs would make the header very long whilst
keeping them is clearly neater... but it leaves a lot of room for
ambiguity and has been pointed out Profile doesn't necessarily fix that.

Hmmm... so from my personal point of view, an ATOM based or ATOM-like
solution seems to be the way to go. A central repository of link types
that can be used across HTML, ATOM, HTTP and more - and one that can be
extended without a standardisation process - seems a friendly way
forward. I guess I have in mind a form on w3.org where you can upload a
new term (subject to checks that every other uploaded word isn't Rolex
or Viagra). I know this is against the ethos of the standardisation
process, but it seems more flexible. GRDDL and POWDER won't be the last
groups to want a new rel type.

On implementation, I've posted more on this issue on this list before
now [5] which includes links to the ICRA implementation of Link Header
which in turn is based on Perl's LWP module and the ease of configuring
Apache and IIS to include the headers.

Phil.

[1] http://lists.w3.org/Archives/Public/ietf-http-wg/2008JanMar/0444.html
[2] http://esw.w3.org/topic/LinkHeader
[3] http://www.w3.org/TR/powder-dr/#assoc-linking
[4] http://www.w3.org/PICS/
[5] http://lists.w3.org/Archives/Public/ietf-http-wg/2007OctDec/0064.html

-- 
Phil Archer
Chief Technical Officer,
Family Online Safety Institute
w. http://www.fosi.org/people/philarcher/

Received on Thursday, 13 March 2008 16:37:12 UTC