Re: What would break? Re: httpRange-14

On 27 Mar 2012, at 06:45, Pat Hayes wrote:
> On Mar 26, 2012, at 12:27 PM, Tim Berners-Lee wrote:
>> The SWAP project, CWM has that built in -- the URI of the document
>> something was read from is kept in the quad store as provenance
>> for every triple read in.  
> 
> OK, but that does not mean that this URI actually denotes that document in RDF. Maybe the semantics of the fourth field in a quad store is different from that of fields in an RDF triple. (In fact, I think we are goingto *have* to assume it is different in order to give any kind of coherent semantics for quad stores.)
> 
> The acid test would be whether you use this URI *inside an RDF triple* with this 'document provenance' meaning.


I was thinking something similar. Using the URI of the resource from which you gleaned some RDF as the identifier for a graph in a quad store is at best a convenient fiction. If you were to ever add statements about that graph to actually record things such as when you retrieved it, which user agent string you used, which Accept header you used, what processing you put it through which which version of your software and so on -- vital statements about the provenance of the RDF that you've gleaned, I would have thought -- then really you would have to use a separate identifier for the graph than for the resource on the web.

But, just like using a URI to refer to a Person or Organisation, reusing the URI of a resource for a graph is a convenient thing to do. In my opinion, we need to recognise that people will do the convenient thing and (a) have a protocol that is flexible enough to cope when they do so (call this error recovery if you like) and (b) provide understandable routes from the convenient path to a more rigorous one.

Cheers,

Jeni
-- 
Jeni Tennison
http://www.jenitennison.com

Received on Tuesday, 27 March 2012 08:45:17 UTC