Re: GRDDL Going to Last Call

>From www-archive as I am alerting Chairs and other relevant people that
we are moving closer to Last Call...somehow this became an argument
about typing, since Ben implied that typing was needed for compability
with RDFa.

Ben Adida wrote:
> Harry Halpin wrote:
>   
>> No, you do not need a complete solution but you need a solid proposal,
>> or you  need to find someone else to help you with the proposal. Again,
>> I *like* the general idea I'm just rather busy myself right now and
>> haven't thought it through.
>>
>> If you can get (or others in SWD) can get a solid proposal through by
>> before our schedule says we should go to CR. Perhaps you could ask Fabien.
>>
>> I do not think that this is too much to ask, and if you are overburdened
>> someone else from SWD WG could take this on. Again, we have done
>> everything requested in order to remain compatible with RDFa, but people
>> interested in RDFa also have to help out.
>>     
>
> Your description makes it sound like I simply waved my hand and said
> "please do this." I tried to have this discussion during two GRDDL
> telecons. I wrote up the need for this in detail in September [1]. I
> wrote an hGRDDL transform in Javascript and a wiki page describing how
> things would work ideally, in December [2].
>   
Ben,

    I'm moving this off the public list www-archive@w3.org and to public
list public-grddl-wg@w3.org. I think your message [1] was very close to
a use-case for typing but not exactly, and writing it up and asking
Fabien to put it in the use-case document would be fine.  Yet currently
the substance of your message [1] was addressed, GRDDL can output
XHTML+RDFa as you specified: "GRDDL transform that can output XHTML+RDFa
as its serialization of RDF".

 What is underspecified is exactly why typing is needed: "ideally, this
would be indicated right when the GRDDL hook is declared, so that a
browser knows ahead of time what the GRDDL transform will output." The
response so far has been that the GRDDL-aware agent could use
media-types or  sniff the output documents types in order to discover
which transform produced "XHTML+RDFa" and then use that. I agree that
could result in unnecessary transforms being run but I don't see how it
breaks anything or that use-case described in [1]. I could see using a
type attribute possibly working (which I think would be better than
namespacing the transformation link, since that would complicate HTML,
but rel element's have type attributes in HTML[2]).

    The only problem with a type attribute is that it 1) assumes
XHTML+RDFa gets a content type (which I think is reasonable, and I hope
in the SWD's path) and that 2) the type attribute normally specifies
"This attribute gives an advisory hint as to the content type of the
content available at the link target address", i.e. the content would be
the media type of transform, i.e. XSLT or  Javascript, *not* the result
of the transform.

    Another option you mentioned is using a namespace in the attribute
content, such as <link rel="grddl:transformation"> or <link
rel="rdfa:transformation">, right? But that's using QNames in  attribute
content, which is broken according to WebArch [3].

    This leaves one typing the rel attribute, which is <link
rdfa:rel="transformation" />, which would force implementers to change
how they detected rel attributes. And it's still unclear what namespace
"rdfa" would refer to.

    Perhaps the final option I can think of is to use a "media"
attribute, but I'm not sure if this is right. Media attributes[4]
usually do not take media types, but media-descriptors. These are
extensible and it may or may not be broken to add media attributes to them.

     If you want to, you could take an action to you or Fabien at the
next telecon to do write this up, or wait till after Last Call if you or
others who want this are busy right now. Again, I think you did talk
about typing transforms and there was some lack of understanding from
various WG members (including myself) about why this was needed and how
it should be done, so it's clear it needs to be written up clearly and
e-mailed to public-grddl-wg@w3.org and then argued out in a telecon. But
without a clear method and accompnay text we can get consensus on, I
can't see what the rest of the WG is supposed to do.

     In fact, the time spent typing this last few e-mails out would
probably have been sufficient to write up the use-case and the proposed
text needed for adding typing.

[1] http://lists.w3.org/Archives/Public/public-grddl-wg/2006Sep/0073
[2] http://www.w3.org/TR/html401/struct/links.html#edef-LINK
[3] http://www.w3.org/TR/webarch/#xml-qnames
[4] http://www.w3.org/TR/html401/present/styles.html#adef-media
> One specific option I brought up was the idea of subclassing
> grddl:transformation. Someone else brought up the idea of using the
> @TYPE attribute on the LINK element. I'm not sure what the best approach
> is, but I do know that the discussion quickly turned to "let's just have
> the client try everything and graph merge."
>
> The GRDDL WG has focused so far on publishing something quickly for the
> well understood use cases. I don't fault you one bit for doing this:
> GRDDL needs to be published ASAP as it's quite useful and it's been
> sitting around as a Note for too long.
>
> However, we shouldn't act like everything was done to coordinate with
> RDFa: I've been raising this issue for a while and the reaction from the
> WG has been lukewarm at best. I understand that different WGs have
> different priorities, so it's fine, that's the way life goes. I'm simply
> giving you my feedback regarding the GRDDL/RDFa coordination. I don't
> think the current spec will be enough for RDFa's needs. We'll do our
> best to fit within the current GRDDL spec, but it would be a heck of a
> lot easier if the transformations were typed.
>
> -Ben
>
> [1] http://lists.w3.org/Archives/Public/public-grddl-wg/2006Sep/0073
> [2] http://www.w3.org/2006/07/SWD/wiki/hGRDDL_Example
>   


-- 
		-harry

Harry Halpin,  University of Edinburgh 
http://www.ibiblio.org/hhalpin 6B522426

Received on Tuesday, 6 February 2007 02:36:05 UTC