Re: RDF/XML and named graphs

On Dec 13, 2007, at 2:08 PM, Jeremy Carroll wrote:

> Alan Ruttenberg wrote:
>> I was thinking less of whether it would be considered legal by  
>> parsers, but whether the semantics would be corrupted.
>> Initially you have triples segregated into graphs, but for an old  
>> parser you have a single graph and some extra triples.
>
>
>> Could there be nasty misunderstandings?
>
> I think there could be, and one would need to be careful.

OK. I was thinking about how such a feature would be advertised, and  
for what uses recommended. Regarding the semantics, I'm thinking  
about those uses of named graphs where the graphs are used e.g., to  
segregate instructions some sort of conditional action. In this case  
the merge would lose this (and possibly have a system act  
inappropriately because of this). Of course one can't formalize this  
meaning for graphs, but that doesn't mean people won't use such  
constructs.

I think, if such a modification were made, that one would want to  
give guidelines including things that were "safe" in some way, and  
plausible demonstrations of some things that were not.

Regards,
Alan

>
>
> In RDF as in the recommendations, it is fairly hard to have  
> different semantic force. This can be achieved by using separate  
> documents. For example, the RDF Test Cases, are described with a  
> Manifest file (which is in RDF), and is intended to be read as an  
> assertion of fact, or intent, or even as a performative "we hereby  
> bless the following tests" by the RDF Core WG, and the W3C. The  
> data files for the tests appear in a similar way on the Web server,  
> and are intended to be read as examples, without any intent of  
> their meaning being taken seriously.
>
> If the author of the triples means all of them: i.e. they are all  
> true in the author's world view, then the motive for separating  
> them into different graphs maybe to make them easier to read or  
> process - so for example metadata may be best kept separate from  
> data or metametadata.
> But if someone else reads it all at a single level, it's not that  
> anything is false ....
>
>
> Jeremy
>

Received on Sunday, 16 December 2007 00:26:09 UTC