W3C home > Mailing lists > Public > www-rdf-interest@w3.org > February 2004

Re: Graph naming?

From: Alberto Reggiori <alberto@asemantics.com>
Date: Wed, 25 Feb 2004 16:11:45 +0100
Message-Id: <ECB02F81-67A4-11D8-8ADC-0003939CA324@asemantics.com>
Cc: "rdf-interest" <www-rdf-interest@w3.org>
To: "Eric Jain" <Eric.Jain@isb-sib.ch>


On Feb 25, 2004, at 11:28 AM, Eric Jain wrote:

>
>> rdfstore:context properties to flag each description block
>
> This is an excellent solution, wish I had thought about it earlier!

this is remains "a" solution anyway and is not "the" solution - even 
though we recon it is not complete and it still needs some work - but 
as far as our application domains are concerned we have found such a 
solution practical enough, working well and also being a fair 
architectural compromise. And it even works for 3rd-part XML and RDF 
tools - which is not a secondary point. But I guess there are people 
having different/more requirements about this "naming" thingie - also 
because such an "artifact" still means different things to different 
people.

More, the property name (context) we picked up could be a bit 
misleading for somebody - perhaps rdf:name rdf:nameID, rdf:Graph 
(rdf:domain, rdf:domainID, rdf:Domain) or similars might be more 
appropriate, "human-friendly" and less messy terminology wise.

But a part the syntactic sugar one wants to use to express such an 
"orthogonal" piece of information, we are convinced it is something 
fundamental for the savvy of the whole semantic web and RDF application 
domain and its deployment.

>
> The only drawback I see is that this solution can end up being a bit
> verbose, but then again it's completely backard-compatible.

right - this is what some more "global" rdf:about attribute might do at 
the RDF/XML document level (like Graham et al. has been proposing); 
and/or in some manifest/collection somewhere.

One possible workaround we are still puzzling is the possibility to use 
some special "collection" contructs like rdfstore:Graph typedNode with 
some rdfstore:nodes  property to inline "graphlets" into main rdf:RDF 
element i.e.

<rdf:RDF.....>

<rdfstore:Graph rdf:nodeID="mine">
	<dc:source rdf:resource="http://mine.trustedsource.com"/>
	<rdfstore:nodes rdf:parseType="Collection">
		<!-- my RDF/XML descriptions go here.... -->
	</rdfstore:nodes>
</rdfstore:Graph>

<rdfstore:Graph rdf:nodeID="your">
	<dc:source rdf:resource="http://your.trustedsource.com"/>
	<rdfstore:nodes rdf:parseType="Collection">
		<!-- your RDF/XML descriptions go here.... -->
	</rdfstore:nodes>
</rdfstore:Graph>

</rdf:RDF>

which is not far from what TriX does of course but it still it looks 
much more like "normal" (standard) RDF/XML - ugly though! but again is 
about syntactic sugar and/or processing software compatibility issues. 
And I am sure we can think of another 50 other possibilities for 
this....

But even so, we should try to avoid to extend current RDF/XML and make 
RDF  as much XML "friendly" as possible i.e. do not add too much noise 
to it - and let people kind of think of RDF like some XML++ vs. brain 
washing them :)

>
> So, if the official serialization syntax is ever revised, defining an
> rdf:context property would in my opinion to be a better approach than
> anything that requires tools to be rewritten from scratch.

well this seems a kind of "big" task, but we are sure it will happen as 
soon as more RDF applications will come to the market - and they will 
need to exchange and merge different pieces of RDF out there.

cheers

Alberto
Received on Wednesday, 25 February 2004 10:11:35 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:14:58 UTC