HTTP Link, rel="powder" cf. rel="describedby"

I had a (very good) meeting with Mark Nottingham yesterday. He's Yahoo's 
point man at the IETF and the person behind the HTTP Link Header 
internet draft to which our DR doc refers [1].

He seems happy that there is a good deal of consensus around most of the 
important issues that draft raises. The area where there is currently 
less consensus is how various relationship types should be registered 
and maintained. This goes way beyond POWDER in terms of scope but it 
clearly affects us since our docs talk about using rel="powder." We've 
followed the current recommendations by using a profile document.

Bottom line - with a bit of political negotiation - Mark believes that 
HTTP link will be able to progress along the route to RFC within the 
time line we need.

But... whilst our use of HTTP Link is right in Mark's view, the 
registration of rel="powder" probably isn't. Section 4.2 [2] of the 
draft says:

"A Link relation is a way of indicating the semantics of a link.  Link
relations are not format-specific, and MUST NOT specify a particular
format or media type that they are to be used with."

I was concerned about this since rel="powder" /does/ indicate a 
particular format (i.e. POWDER). I raised this on the HTTP list and 
Jonathan Rees replied [3] that he thought this referred to the origin of 
the link, not its target. Mark said no - actually the intention is that 
/neither/ end of the link should be format-specific - that's the job of 
the MIME type.

I said that we were wary of trying to register a new MIME type - after 
all, POWDER is either XML or RDF/OWL (semantic extension 
notwithstanding) and that HTML Profile meant we didn't /need/ to 
register either rel="powder" or a new MIME type. Well... that's true but 
we are talking about registering the @rel type so that argument rather 
loses potency!

Mark pointed me to a doc [4] that is an entry point for a description of 
how we would register the POWDER Media type which actually looks pretty 
simple - being in a W3C Rec document means that IETF is likely to agree 
to the new type with little delay.

To get to the point, Mark's recommendation is that we

1. Use a more generic @rel type of describedby (something other groups 
want as well btw)

2. Register a POWDER-specific Media type. I guess ours would be

application/powder+xml

and

application/powder-s+xml

???

Neither of these registration steps is particularly hard to do.

In terms of the WG's process, I suggest we teat this as a Last Call 
comment and deal with it when we resume in September - *unless* - Matt - 
  you advise that /if/ we were to make such a change we'd require a new 
LC version, in which case we may need to take a couple of resolutions by 
e-mail before those docs you're working on are fully published (er, 
which I believe is scheduled for a week today)

Phil.



[1] http://tools.ietf.org/html/draft-nottingham-http-link-header-02
[2] 
http://tools.ietf.org/html/draft-nottingham-http-link-header-02#section-4.2
[3] http://lists.w3.org/Archives/Public/ietf-http-wg/2008JulSep/0122.html
[4] http://www.w3.org/2002/06/registering-mediatype

Received on Friday, 1 August 2008 08:22:10 UTC