Re: [RDFa] The CLASS attribute

At 4:36 PM +0100 14 02 2007, Simone Onofri wrote:
>So, Using class with semantic meaning, this overload the current use
>of @class. Well:
>
>- Use for style in CSS
>- Use for semantics in Microformats (and also eRDF and more)
>
>So, if we use on a single page also more dialects we can have:
>
>... class="mystile first-dialect second-dialect" ...

The counter-argument is that the style key is broken if it is not
semantic in the first place.

Use @class for category membership; that's a clear and simple
concept.

I'm afraid that this is one where I have to side with Tantek and
respectfully disagree with Steven and Simone.

All styles reflect some under-articulated semantic.  "For style"
does not excuse authors from articulating the semantics better,
as that is where the semantic rubber hits the road.

We had a joint PFWG:WCAG telecon a while back where the then
activity lead for SW assured us that there were multiple good
options for binding out-of-document RDF metadata explaining
the dialects to tokens used in @class in an HTML 4.01 instance
document.  So there is no excuse for an RDF - breathing processor
to ever be confused about tokens that have explanations in
different dialect-references.

>and this is not clear also for humans and create more confusion also
>for machines.

For semantics, we need a place where the aspects come together.
There is no stable factorization beyond a few clear foci and we
don't have them teased out yet.

If the machines are based on a tuple-space library for selectors,
there is no machine confusion.

One surface style selector will rationally key off several 
backoffice-semantics aspects.
You can't put a dividing line between style selectors and semantics
or you have thrown the baby out with the bathwater.

>It's a good idea use only "namespaced" class but if we can use a
>specific @attribute for semantic layes, this should be the best
>choice.

There are no semantic layers; there are semantics in all layers. The
layering is driven by performance in the delivery pipeline; we won't
get a semantic layering to take hold, ever, because the semantic
space is not layered; it can be provisionally aspect-factored, but
even that is not forever. [other than hard-to-process RDF glosses
that few process and hence lack the viral energy for maintenance of
interoperable concepts in "dialects" or patches of related concepts.

Now, to back  off from the pro-microformats blast:

In WAI-ARIA we *are* as of now using @role for things that have
reference explanations that are accessible by QName.  But this is
really because we took what HTML WG offered us as their choice;
not because we did a conclusive, independent analysis of where to put
what metadata.

Al

>Simone
>
>>On Wed, 14 Feb 2007 14:39:16 +0100, Ben Adida <ben@adida.net> wrote:
>>
>>>  Ivan Herman wrote:
>>>>  +1
>>>>
>>>>  Steven Pemberton wrote:
>>>>>  I want to reraise my position: leave class alone, and use something new
>>>>>  for what we want. I still have the feeling that @role can do the job.
>>>
>>>  There is a *huge* lost opportunity if we do this, which is that we don't
>>>  show a natural progression from microformats to RDFa. Microformats
>>>  already use the CLASS attribute for this kind of thing, and it's a well
>>>  accepted practice. New attributes should be used sparingly.
>>>
>>>  -Ben

Received on Wednesday, 14 February 2007 18:55:06 UTC