Re: GRDDL notes from Liam Quinn

On Tue, 2006-08-15 at 23:20 -0400, Harry Halpin wrote:
[...]
>   ---------- Forwarded message ----------
> Date: Tue, 15 Aug 2006 20:23:15 -0400
> From: Liam Quin <liam@w3.org>
> To: Harry Halpin <hhalpin@ibiblio.org>
> Subject: Re: GRDDL quesion
> 
> 
> GRDDL's not really a triplet.. it's
> 
> 1. the HTML document in which the GRDDL appears
> 2. a pointer to a transformation, e.g. in XSLT
> 3. the internet media type of that transformation (e.g. XQuery, XSLT)
> 4. the result of the document (RDF)
> 5. and what is described by that result (this could be implicit)
> 
> Let's ignore (5) for now.
> 
> Since GRDDL rams everything into a triplet, how do I say that I'd
> rather use XQuery than XSLT (say) if both are available?

In the usual way, no? i.e. you give the XQuery media
type preference in the Accept: header when you follow
the link to 2.

> And what if I fetch the resource and it's not XSLT at all, but
> XQuery?

Well, what if you prefer .png but you get .gif or .svg back?
Such is life, i.e. such is the HTTP protocol. If the transformation
provider only gives XQuery and you don't grok,
you're going to lose. That's why the GRDDL spec advises
transformation providers to stick to what's widely deployed:

 "We expect most consumers to support XSLT version 1[XSLT1] for the
foreseeable future, though XSLT2[XSLT2] deployment is increasing."
 -- http://www.w3.org/2004/01/rdxh/spec#txforms

Perhaps it's worth mentioning XQuery there, though I haven't
seen anybody implement support for it.


>   In other words, (3) is implicit in the relation name,

I don't understand where he gets that impression.

> but we shouldn't be dictating the media type of a resource in
> the source, we don't say
>      <a href="foo" type="text/html" ...
> as that would prevent the server from deciding based on content
> negotiation, would prevent upgrades to XHTML, etc.

Yes, we do; that is: we're allowed to. I much prefer
links that are more generic, but sometimes a specific
link to the .xsl is in order. This is completely vanilla
web architecture; GRDDL is totally orthogonal.

It's just like a stylesheet link; the link syntax allows
you to give a generic link:
  <link rel="stylesheet" href="my-style" />
or a more specific link...
  <link rel="stylesheet" href="my-style.css" />
or even to pre-judge what media type you're going to get back:
  <link rel="stylesheet" type="text/css" href="my-style.css" />

The author gets to choose what sort of link to express.

> Of course, if the transformation is done on the server (as I
> think it should be) then one could even link to a static RDF file.

If the transformation is done on the server, then it's just RDF/XML,
to the consumer; the difference between a static file and running
some code is invisible. But if every producer were willing to publish
RDF/XML, we wouldn't need GRDDL in the 1st place.

> Liam
> 
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Wednesday, 16 August 2006 06:58:19 UTC