Re: Named graphs etc

On Mar 11, 2004, at 11:28, ext Jeremy Carroll wrote:

>
>
> I think we are agreed that:
>
>  the choice between an asserted attribute in the XML and vocab in the 
> triples is an important one
>
>  that it is not obvious which way to go

I'm myself pretty much sold on the vocabulary approach.

>
>  (Jeremy, Pat and Patrick have all argued for an asserted attribute at 
> some point)
>
> Judging by Pat's last message, at least for now, we are trying a vocab 
> solution. I need to read and probably reply separately to that 
> message.
>
>   The obstacle of legacy RDF/XML that does not have such an attribute, 
> and legacy RDF/XML tools that do not support such an attribute is one 
> reason for using vocab, but we should try and enumerate the pros and 
> cons.
>
>
> Pro XML attribute
>
>   - very clear
>   - publisher has clear control
>   - easy solution to bootstrapping problem
>
> Con XML attribute
>
>   - publisher's identity left implicit

Could add yet another attribute to indicate authority. And another
for a signature...

>   - legacy

I think this is the particular nail in the coffin for this approach.

>   - third party assertions impossible

This is a *PRO*, not a con!  ;-)

I think that assertion and agreement are two very different things.

Just because I agree with the statements in some graph does not
mean that I assert that graph (even if I myself assert another graph
having the same statements).

Eh?

I wonder if what you really mean by rdfx:assertedBy is really more
like rdfx:affirmedBy ...?


>   - adds additional theoretical level, isn't assertional status just 
> more metadata?

I think we add that level for either the XML or vocab approach.

>
>
> Pro Vocab
>
>   - uses existing mechanisms, and hence partially addresses legacy
>   - enables third party assertions,

Not necessarily, if the interpretation of statements employing that
vocabulary to qualify a graph is limited to qualifying the graph
in which they occur.

Perhaps we need to split this up into Intra-Graph Vocab and
Inter-Graph Vocab, the former providing for self qualifying
graphs, and the latter allowing for one graph to have statements
qualifying another.

> or single assertion by author for many docments
>   - can derive assertional status, so system is extensible by users

I'm not sure how you could reliably derive the intent of the author
as to whether a set of statements was intended to be asserted or not.

>   - vocab can require publisher to be identified by a URI node

Exactly how does one do this in RDF? I don't recall any facility
for differentiating between URI denoted resources and bnode denoted
resources in the range of a property. Does OWL provide such a
distinction?

This would, though, be a pro for the XML approach, in that one could
have an attribute for the authority that required URI values.

> (well I guess a bnode would do, but I wouldn't trust it!)

Nor I, unless there was some owl:sameAs equating it to a URI that
I trusted...

>
> Con Vocab
>
>   - can derive assertional status so bootstrapping is complicated

I would argue that this is a con for Inter-Graph vocab but not
Intra-Graph vocab.

>   - bootstrapping is logically twisted, since we seem to need to 
> assume that a graph is asserted in order to examine its assertional 
> status

One person's 'twisted' is another's 'elegant' ;-)

>
>
> What have I forgotten? Short statements please, no advocacy.

I tried... really I did...

Patrick


>
> Jeremy
>
>

--

Patrick Stickler
Nokia, Finland
patrick.stickler@nokia.com

Received on Thursday, 11 March 2004 07:13:59 UTC