Re: Information resources and RDFa

Ivan,

There is no need to apologize: your input on this is important, and  
we encourage more chiming in.

We have explored the idea of nested statements to build deep RDF  
graphs. None of the solutions we've come up with have been very  
pretty in the end, because they tend to generate vastly more triples  
than the author intends. We're open to continued suggestions on this,  
of course.

Note that my last email wasn't intended to say that we won't have  
bnodes, just to say that it's okay to not try to build bnodes  
everywhere, because I believe that a lot of uses of RDFa will not be  
bnode-dependent. But surely, there will be some bnode-dependent  
implementations.

Our current bnode support is best when using existing HTML elements  
without IDs. For example, your example could be written as:

=========

<body>
...

<p>
	<link rev="dc:creator" href="" />
	blah blah blah, <meta property="dc:name">Ivan Herman</meta>
</p>

...
</body>

=========

Now, your examples below are interesting, in that you're only asking  
for special rules when inside a LINK or META. Note one problem: by  
default those will be display:none, and if you choose to display  
them, then everything inside LINK will be clickable.... So I'm not  
sure that's really what you want. The above seems more along the  
lines of what you're seeking.

What do you think?

-Ben

PS: Keep an eye out for my next email, which *might* simplify things  
a bit more, if the TF agrees with my note.


On May 11, 2006, at 7:24 AM, Ivan Herman wrote:

> Ben,
>
> I am sorry to chime in this mail, and maybe the issue I raise have
> already been discussed before...
>
> However, I respectfully disagree on your remark below on the usage of
> bnodes. At some point a few months ago Steven Pemberton and I played
> with the convertsion of a set of RDF statements (on public
> presentations, the RDF data that I use for our tals database[1])  
> and one
> of the difficulties was the fact that we had to *name* each node, even
> if they were bnodes. Not only it is a pain in the backside because one
> has to be careful to use unique names, but that could make any  
> automatic
> generation of such files more difficult.
>
> Looking at Mark's example (for example, [2]) I wonder whether it is
> possible (ie, whether it crashes with something else) to push  
> nesting a
> little bit further and say:
>
> <head>
> 	<meta>
> 		<link rel="dc:creator" ref="http://blabla/>
> 		...
> 	</meta>
> </head>
>
> or having the same possibility to do something like that in the body:
>
> <body>
> 	<link rel="dc:creator">
> 		<p>Bla bla bla <span property="dc:name">Ivan Herman</span>....</p>
> 	</link>
>
>
> Both cases would be the equivalents of using '[ .... ]' in turtle, ie,
> the creation of a blank nodes without bothering with the name...
>
> If that was considered before and shot down with good reasons, then I
> stand corrected:-)
>
> Ivan
>
> [1] http://www.w3.org/2004/08/W3CTalks
> [2] http://skimstone.x-port.net/node/272
>
> Ben Adida wrote:
>>
>> [snip]
>> Steven, I'm fairly certain that Mark meant href="#author", and not
>> href="#about". Mark, can you confirm? I also don't think that's a
>> bnode... quite the contrary, it's an addressable node, it's just not
>> an identified XHTML element. In fact, I suspect more and more that
>> users of RDFa will probably use bnodes rarely, because it's just much
>> easier to identify all the nodes when you're dealing in HTML.
>
>
> -- 
>
> Ivan Herman
> W3C Communications Team, Head of Offices
> C/o W3C Benelux Office at CWI, Kruislaan 413
> 1098SJ Amsterdam, The Netherlands
> tel: +31-20-5924163; mobile: +31-641044153;
> URL: http://www.w3.org/People/Ivan/

Received on Saturday, 13 May 2006 16:42:18 UTC