Re: Determination of subjects/objects

Indeed, Ben, that is what I meant....

I do _not_ want to enter the debate whether we need bnodes or not. I do
believe we need it (though we should not abuse them) but that _is_
another discussion altogether.

At the moment, in RDFa, I see two ways to generate a BNode:

- use the CURIE notation (this thing should be written down!), ie, if I say

<span rel="a:b" resource="[_:123]">
 <span property="q:r">yep</span>
</span>

this means

<> a:b _:123.
_:123 q:r "yep".

In turtle and this is a _named_ bnode. I can of course reuse _:123 in
the RDFa file somewhere else. But copy/paste this portion from one file
to the other is inherently dangerous!

- using our current set of rules I can say:

<span rel="a:b" instanceof="foaf:Person">
 <span property="q:r">yep</span>
</span>

which will generate a _typed_ bnode already, meaning

<> a:b [ rdf:type foaf:Person; q:r "yep". ].

Which is very different then before: I _can_ cut and paste it without
danger, but it forces a type on the bnode.

What I simply do not see at the moment is how I can use RDFa to generate
something like:

<> a:b [ q:r "yep" ].

Ie an unnamed and untyped BNode. There are lots of examples of using
such constructs.

My solution is to say:

<span rel="a:b" resource="[_:]">
 <span property="q:r">yep</span>
</span>

I am not sure that a resource="[_:]" is much more complex or esoteric
than resource="[_:123]". In some ways, due to the cut/paste issues, I
find the latter more dangerous and maybe even less useful. But I do not
see any other way out of this requirement:-(


B.t.w., my initial thought was to say

<span rel="a:b" resource="">
 <span property="q:r">yep</span>
</span>

which would look more natural. Unfortunately, it is wrong, because
href="" refers to the 'top', and I do not think we should have a
different interpretation for resource...

Ivan

Ben Adida wrote:
> 
> So, Mark, correct me if I'm wrong, but I believe we have the ability to
> name bnodes using the CURIE notation, e.g. resource="[_:author1]" and
> the like. That hasn't changed, as far as I know.
> 
> The question on the table was whether we could use "_:" as an explicit
> request to generate a new unnamed bnode.
> 
> Ivan, did I misunderstand?
> 
> -Ben
> 
> Knud Hinnerk Möller wrote:
>> Ok, here is my use case, if I remember correctly:
>>
>> - I generate RDFa (using the SWRC ontology) from BibTeX
>> - a paper has multiple authors
>> - there are many ways to represent that in RDF
>> - SWRC just uses multiple (book swrc:author aFOAFPerson) statements for that
>> - however, that doesn't keep the order of the authors
>> - that's why I _also_ use one (book swrc_ext:authorList
>> sequenceOfAuthors) statement
>> - I do both because I want to stay compatible to the original SWRC, but
>> also give the possibility of getting the author order
>> - of course the foaf:Person resources in both approaches should be the same
>> - I don't have URIs for the authors, so I need to use bnodes (or make up
>> URIs, which I don't want to do)
>> - because I want to refer to the same resource through both swrc:author
>> and swrc_ext:authorList, I need to be able to name the bnodes within
>> that graph
>> - i.e., I need named bnodes
>>
>> I hope that makes sense. :)
>>
>> Cheers,
>> Knud
>>
>>
> 
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf