Re: XHTML-RDFa draft made public

Mark Birbeck wrote:
> 
> Hi Ivan,
> 
> Thanks for the comments.
> 
>> my apologies for chiming in here, although I am not part of the HTML WG.
>> But I do follow the RDFa work very closely....
> 
> :) No need to apologise!
> 
> 
>> I am not sure what the role of this document really is. *If* it is (part
>> of) the technical specification of RDFa, then I am a bit surprised by
>> some facts.
> 
> It was originally intended that this document simply describe the
> schemas and DTDs for the RDFa module, but it became clear that it was
> difficult to see the schemas in isolation from RDFa. That's not to say
> there isn't more work that can be done though, on getting the format
> and content right in relation to other RDFa documents.
> 
> 
>> For example:
>>
>> - Section 3: the role of the hreflang, hrefmedia, hreftype attributes.
>> These have never been discussed in terms of RDFa lately, I am not sure
>> what their meaning is in terms of RDF (and I am not sure they should
>> have any meaning!). [To avoid any misunderstandings: in abstract these
>> do make sense, it is the context that bothers me!]
> 
> Yes, it's true that these attributes weren't discussed in the context
> of RDFa, but the problem we had is that in XHTML, the attributes
> relating to linking come as a package. It was therefore felt that they
> should be handled together.
> 
> Also, if you look at RDFa as not just a way of expressing triples in
> mark-up (the RDF side) but also as a way of defining the triples that
> are generated by arbitrary mark-up (the HTML side), then you could
> argue that we need to say something about pretty much all HTML
> constructs (or at least those that relate to linking).
> 
> So I sympathise with your shock. :) And I think we should put these
> attributes onto the agenda for discussion, and see what people think.
> 

I could very well imagine to cut these things into two. Have an RDFa
module that would be the strict minimum for the expression of RDF in
HTML, so to say, and a separate module, fully under the control of HTML,
that would also provide some predefined RDF triplets in terms of RDFa,
for example. Something like that...


> 
>> - Section 4: some of the examples listed here seem to go beyond the
>> current content model of HTML, namely the fact that a meta element can
>> have a child, that it refers to the element being the child of 'link'.
> 
> Sure, but we can do that in the module, if we agree that we want to.
> We're not removing any syntax from HTML, we're augmenting it.
> 
> Equally, we can reduce the level of change if we want to, as well--but
> the main point I'm making is that this is not an XHTML 2 v. XHTML 1.x
> question, it's just a question about how much or little we want to put
> in.
> 

I am not sure I understand your point here, Mark. XHTML2 plans to change
the HTML content model, XHTML1.x does not (at least in my
understanding). If RDFa goes in direction of XHTML1.x, then this is an
issue. [Whether 'change' means 'augment', is a bit besides the point in
this respect.]

> 
>> It also refers to a list of predefined values in the XHTML2 namespace.
>> The current direction of RDFa is *not* to bind it to XHTML2; this is a
>> crucial point in the deployment strategy of RDFa
> 
> That's probably a cut-and-paste error. I'd like to stress that we're
> pretty clear on our _goals_ with this module.
> 

O.k.

> 
>> - I also have the impression that we got to a nice equilibrium point on
>> the literal datatype discussion on the mailing list (see the thread
>> starting at [1]), but this is not reflected in the document (maybe I
>> have just missed it).
> 
> There has been a lot of discussion on this in two or three subsequent
> telecons, and the equilibrium that was reached was unfortunately a
> negative one; that we would lose something whichever side we came down
> on. We were therefore waiting for an action item from Elias to be
> completed, which would identify what the problems with the _current_
> solution are from the point of view of triple stores, and then
> hopefully we can wrap this up one way or the other.
> 

Oh, not again...:-) I really liked what we ended up with, it looked like
a perfect base for a consensus...

> 
>> *If* the goal is to have a meta module for XHTML2, that is fine with me
>> and I am not at my place to make any comments (although the document
>> should make it very clear the context of this document). But if the goal
>> is to issue an *XHTML1* module for RDFa with the minimal possible
>> changes and with no dependency on XHTML2 then I fear this document will
>> raise some eyebrows...
> 
> I hope I've answered each point, but I'll just recap in answer to
> these final two points you raise, so that we're sure (and in order to
> reassure :):
> 
> 1. The goal is not to have a metamodule for XHTML 2...or to be more
> precise, the goal is not to have a module that is _specific_ to XHTML
> 2. Since XHTML 2 could be built with XHTML M12N (perhaps an updated
> version) then we hope that the module we are producing now can be
> re-used in the future, and serve both XHTML 1.x _and_ XHTML 2.
> 

Yep, that is fine.

> 2. If I've understood all of your points, the eyebrow raising would
> concern (a) the predefined values from XHTML 2, which are probably a
> cut-and-paste error, and (b) the attributes like hreflang, which do
> need to be discussed, even if we decide to drop them from this
> document.
> 

Beyond this,  I think the extension/change of the HTML content model
(children of link) would be a problem, too. If we want to have RDFa
accepted by the HTML working group, we have much more hope to do that if
the RDFa extensions are restricted to the introduction of some new
attributes. I would definitely vote against any change on the content model.

> 
>> My concern is, again, the wide adoption of RDFa
>> ASAP, that is why I allowed myself to chime in.
> 
> Chiming in is most welcome, but at the same time I hope you see that
> adoption is our concern too, and is of course the very reason why we
> devised the idea of a XHTML 1.x module that makes use of XHTML M12N in
> the first place.
> 

I know. We share the concerns...

Ivan


> Regards,
> 
> Mark
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
URL: http://www.w3.org/People/Ivan/
PGP Key: http://www.cwi.nl/%7Eivan/AboutMe/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf