Re: talking again

>> 	<link about="#some(a)" rel="foaf:knows" href="#some(b)" />
>
> I understand in theory that you can do this, but it is pretty
> misleading. Not least
> since the attribute name "about" is copied from the URI-taking attribute
> of the
> same name in W3C's existing RDF/XML syntax. I'm a standards nerd, and
> I'm
> still fuzzy on what the rules would be for detecting whether the value
> of an
> "about" attribute here would be an xpointer-esque bnode localID, versus
> a (possibly relative) URI reference.
>
> Aside: would xml:base affect this attribute? including the
> xpointer-bnode IDs?

I think you would have to say that the some() XPointer scheme refers to a  
local bnode in the document that uses it, and is therefore independent of  
the LHS of the URI using it. It would be magic in the some() scheme.

>> These *look* like URIs (and have to in order to be able to use the same
>> attributes), but all they really are is a message to the RDF serialiser  
>> to
>> transform them into RDF bnodes.
>
> So what would you write if the RDF triples you were trying to encode
> really
> did want to make assertions about local xpointers?

We're not doing anything special with xpointers, only one scheme. So other  
xpointers would work as they were intended...

> There are legitimate triples that can be written with an xpointer-based
> local URI reference; we just need to take care not to confuse them with
> triples whose subjects are bnodes.

... therefore not a problem.

>
>> 			Although they are URLs in the HTML world,
>> when they get into the RDF world, they are not any more, they are  
>> bnodes.
>>
>> The advantages of this approach:
>>
>> 	The markup is simpler
>
> ...by hiding structure in attribute content conventions, sure.

I disagree. There is no hiding going on: it is an explicit notation, and  
the metadata really *is* 'about' the bnode.

>> 	Less explaining to do
>
> I'm not so sure on that point. Particularly if you include the
> confusion
> it'll create around the other "about" attribute, ie. the one in RDF/XML
> which
> doesn't behave like this.

I think it is easier to say "if you don't know what you are referring to,  
you can give it a name as a placeholder" rather than having to explain  
that there are two new attributes that replace 'about' and 'href' in  
certain cases.

>> 	We only have to argue about one name instead of two :-)
>
> Yup. I think naming these attributes is worth some thought. Hmm...

> ps. congrats on shipping a new WD :)

Thanks... Already 300 new issues come in to the database.

Steven

Received on Thursday, 2 June 2005 12:45:32 UTC