Re: [Proposal] ISSUE-42: How does RDFa deal with @src

Hi Ben,

There are two problems with this. The first is that you might not have
a @rel value on your <img>. So for example, there may be use cases for
the following construct:

  <img src="my-photo.jpg" class="foaf:Image" />

(Especially since everyone is against the idea of generating a triple
for xh:img.)

This means that we can't rely on the presence of @rel or @rev to tell
us how to interpret @class.

The second problem is that as things stand I can't find a
justification for applying @class to @href in the way that you seem to
be suggesting. The only thing I can find that proposes what @class
actually means is in the first link [1] that appears in the page for
the issue itself [2]. The post says:

--- STARTS ---

  <div role="wai:toolbar">

  is equivalent to this:

      <link rel="xh:role" href="[wai:toolbar]" />

  To clarify what @class should do, I'm also suggesting that this:

    <div class="foaf:Person">

  should be equivalent to this:

      <link rel="rdf:type" href="[foaf:Person]" />

--- ENDS ---

Obviously things have changed in relation to '<link> everywhere',
since then, but the general point seems relevant. So to use your

  <a rel="dc:creator" href=""

the proposal above would have it be equivalent to the following
(pretending for a moment that you can have multiple instances of the
same attribute, just to make it easier to write):

    rel="dc:creator" href=""
    rel="rdf:type" href="[foaf:Person]"

I see from your example in a separate thread that when using striping
you've suggested that the @class attribute applies to the 'child'
bnode, but I find that counterintuitive, and don't recall it being
discussed. But either way, I don't see how that applies here, since
striping is for when we have a predicate with no @href, and it would
obviously be meaningless to have an <img> with no @src (which is what
you'd have if you say that @src is the same as @href).

I'm not saying by the way that @src on <img> (and <object>) is a
'special case'. I'm simply saying that it has its own defined rules,
just like <link>, <meta>, @rel and lots of other things have their own
defined rules. The main point I'm driving at is that there doesn't
seem much to generalise between @src and @href.



[1] <>
[2] <>

On 21/06/07, Ben Adida <> wrote:
> Mark Birbeck wrote:
> > Anyone else have a view? In particular are there any use cases for
> > knowing that something is an image, independent of a value in @rel or
> > @rev?
> On this issue, I think I agree with Ivan regarding the xh:img triple: it
> will seem inconsistent. Why IMG and not OBJECT, TABLE (gasp, tables?),
> LINK, SCRIPT, etc..?
> >> Also, are we sure that
> >>
> >> <s> rdfs:label "alt".
> >>
> >> is the best representation for this?
> I don't have a strong opinion on this one.
> > Right. That's fine, and I can go either way. The only thing that I
> > think we do have to discuss though, is @class. My other proposals are
> > about generating 'extra' triples (like 'xyz is an image'), the use of
> > @class on an <img> is something that is already allowed, and we
> > therefore need to provide an interpretation.
> I agree that we need to provide an interpretation for class="xy:zz" on
> an img. I think it's clear that this applies to the IMG, but I think we
> don't need to special-case it too much. We've already set the precedent
> that @REL automatically sets the subject of all contained triples,
> including any triple generated by @CLASS, to be the OBJECT (in that case
> a bnode, but the OBJECT nevertheless). I think we can carry this rule
> over. If there's a REL on an IMG, then @SRC acts like @HREF and becomes
> the subject for all triples generated by contained elements. Thus @CLASS
> is the rdf:type of the @SRC attribute.
> I believe this carries over to the following situation involving an
> anchor rather than an IMG:
> <a rel="dc:creator" href=""
> class="foaf:Person">Ben</a>
> I'm pretty sure that should be interpreted as:
> <> rdf:type foaf:Person .
> right?
> Now, what to do if there is a @REV? I think the same thing applies,
> actually, meaning that @SRC is the subject of the @REV predicate, but
> *also* of the @CLASS rdf:type and all contained elements.
> One last thing: I think all of above should apply to OBJECT/@SRC.
> -Ben

  Mark Birbeck, formsPlayer | +44 (0) 20 7689 9232 |

  standards. innovation.

Received on Thursday, 21 June 2007 20:41:37 UTC