[whatwg] Ghosts from the past and the semantic Web

[Tab Atkins Jr.]

On Fri, Aug 29, 2008 at 3:10 PM, Ben Adida <ben at adida.net> wrote:

> Tab Atkins Jr. wrote:
> >
> > Is this approximately your intended message?  If so, then how do you
> > square this with the plain-to-see usefulness and heavy adoption of CSS?
>
> CSS is great: it actually separates semantics and presentation.
>
> > I also feel the comparison to CSS is quite exact - with CSS you have to
> > first map your html to semantic categories, and then map those
> > categories to presentional settings.  All crdf changes is that you're
> > mapping the semantic categories to external metadata categories.
>
> The repetition of semantic-semantic is exactly why your proposal is not
> nearly as useful as CSS.
>
> If you're going to map your HTML to semantic categories in the first
> place, why not make it the RDF property right away? There isn't nearly
> the same separation of concerns in your proposal that CSS brought to the
> table.


Very good question, and I agree that it's a subtle point.  It seems to be
one that several people on the list agree with at least somewhat, though.
^_^

The issue isn't adding semantics - we do that automatically every day when
we use @class and @id in our code.  The issue is relating these semantics to
an external categorization schema.  This is a more complex issue than simply
adding semantics, as is obvious by the relatively large amount of extra
markup required to insert RDFa into a page.

The other issue is a bit different.  Yes, it's just adding semantics, *but
we do that everyday with @class and @id*.  RDFa is an *additional* semantic
structure that operates independently from our normal semantic system.  If
we're already going to be adding semantics to our data, we might as well
hook any more involved scheme into the existing system if we can.  This is
why people kept asking "Why not just use @class like microformats does?".
We don't want to throw away our existing semantics!  As you (or perhaps
Manu) said before, the problem with using *just* @class is that rdf data
comes in triples, while @class only provides a double (class-name and
content).  This solution extends that class-based double into a full
triple.  And, again, it does so within our existing semantic structure.

I know you think that RDFa is the necessary first step, with alternate
syntaxes like CRDF or similar coming later.  But this isn't 10 years ago.
It's not even 5 years ago.  We web authors *just don't like* pushing extra
data into our pages, especially when we're going to have to repeat the same
data over and over again.  Call it an acquired distaste; regardless, you'll
be fighting this tendency the whole way, unless the person's totally gotten
the religion.  The rest of us who might consider rdf useful but aren't in
love with it will just sieze this as an excuse to toss it away.

Will an easier syntax make us less likely to toss it away?  Hard to say.
However, it *will* reduce the friction you have to fight against before you
run into someone's *real* feelings on rdf.  At the very least, you get to
start the conversation from a better position.

I'm really, honestly not just trying to stall you, or throw more work at you
so you'll go away.  Like I've said, I'm definitely not sold on RDF, but I'm
willing to listen, and some of the applications down the pipeline actually
seem realistic and useful.  I just know how I feel, and I think a lot of
others feel similarly, and I know what would make me feel *better* about
supporting this initiative.

~TJ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.whatwg.org/pipermail/whatwg-whatwg.org/attachments/20080829/669f81a7/attachment-0001.htm>