Re: HTML+RDFa source updated (ISSUE-97, ISSUE-144, ISSUE-146)

On Jan 5, 2013, at 01:18 , Gregg Kellogg <gregg@greggkellogg.net> wrote:

>> On Fri, Jan 4, 2013 at 2:17 PM, Gregg Kellogg <gregg@greggkellogg.net> wrote:
>> 
>> 
>>> The only question I have is: link uses @rel in HTML; is that allowed for a Lite? I would think yes, but this may have to be written down somewhere...
>> 
>> I would say that the use of @rel in <link> is not part of RDFa Lite, if the values of the @rel attribute would be interpreted by an RDFa processor. In other words, they can have terms which are ignored, but not CURIEs or IRIs.
>> 
>> 
>> I am forced to disagree.  We have no way of constraining an RDFa Processor to only do things in a Lite context or a non-Lite context.  Consequently, any occurrence of @rel is going to be interpreted by a conforming processor.  And one conforming Processor cannot work differently than another with regard to the (minimal) triples generated, so all of the values of @rel are going to be processed. 
> 
> You're absolutely right that a conforming processor will interpret @rel as it would be anywhere else; it's just that, as a publishing profile, the use of @rel for RDFa Lite markup is not described, and I don't think it should be in the case of <link> either. Therefore, anyone trying to publish HTML5 with RDFa Lite using the <link> element shouldn't use @rel for RDFa markup; this may mean that @rel is used for other purposes, for which we already have text in RDFa Core to support.
> 
>> In theory a validator could flag the use of @rel on a <link> element, but why?  @rel is legal everywhere according to RDFa Lite.  At least that is my reading of the Profile.
> 
> Ivan may do something like this in his processor, as I believe he outputs warnings when non-RDFa Lite is detected; I don't know what he does in the case of @rel for non-RDFa usage, though.


I *may* output warnings when requested, indeed, but my processing behaviour is not changed.

Actually, at the moment, I skip warning if @content is detected on <meta>, because that is a legitimate attribute for <meta> per HTML. 

But... the HTML Lite rec says explicitly:

[[[
However, even if authorized by the Host Language, the usage of rel and rev should be restricted to non-RDFa usage patterns, as defined by the Host Language.
]]]

I am not sure why we included this, but there we are. Not worth the trouble reopening this issue.

I suspect my implementation should either not issue a warning for link+rel, or should check whether it is an RDFa usage pattern, ie, whether the value of href is a CURIE or a URI. Probably the latter. Wait... using link with a URI is also a legitimate usage pattern! Ie, only CURIE-s are to be flagged I guess...

Ivan

> 
> Gregg
> 
>> -- 
>> Shane P. McCarron
>> Managing Director, Applied Testing and Technology, Inc.
> 


----
Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
FOAF: http://www.ivan-herman.net/foaf.rdf