Re: ISSUE-30: How does SPARQL's notion of RDF dataset relate our notion of multiple graphs?

On Apr 18, 2011, at 9:47 AM, Pierre-Antoine Champin wrote:

> [snip]
> On 04/18/2011 03:40 PM, Pat Hayes wrote:
>>>>>> My very sketchy feeling that if we define a good old class, say,
>>>>>> G-box, we can then:
>>>>>> - say that <g> rdf:type G-box which is the identification of a
>>>>>> g-box
>>>>> This by itself would not attach the name to a particular g-box,
>>>>> however.
>>> neither does
>>> <foaf.rdf#me> a foaf:Person .
>>> attach the name to a particular person.
>> Of course; but in the case of *graph* naming, we expect more than
>> simply having a description; we expect the name to be usable to actually
>> locate and get (a representation of) the graph. So we need a 'baptism'
>> syntax or convention which does the necessary attaching to the name to
>> the graph.
> We are taking about a g-box here, so "something whose state is a graph",
> not the graph itself.

Well, OK, I agree that makes the most sense. We still need a baptism, though.

>>>> Correct. Some hand-waving may be necessary when we define g-*.
>>> well, if the URI of the g-box is in the http: scheme, and it is
>>> dereferenceable (with a 200 OK code), then the HTTP protocol may provide
>>> that hand-waving...
>> Not by itself. We need to actually state (it can be as simple as a
>> statement in the specs) that the http GET is what indeed identifies the
>> graph being named. (And of course this statement needs to be phrased
>> very carefully to allow for g-boxes and so forth; are we naming the box
>> or the graph that is its state at the time of naming? What is GOT
>> (GETted?) by a 404 error or a 303 redirect? And so on.)
> Well, HTTP explicitly states that a URI identifies a resource [1], and
> then gives semantics to the status codes one obtains when sending HTTP
> requests to the given URI.

But it is (notoriously) unclear what exactly that word 'identifies' means, and several prominent Web gurus are on record with the view that whatever it means, it most certainly is not synonymous with 'refers to' or 'denotes' as used by semantic specifications. So it is still up to us to declare what specifies the 'URI is the name of graph' relationship, and how it relates to http. 

> I like to think of 'identification' in HTTP as a subset of
> 'identification' in RDF, more precisely as the restriction of this
> relation to "information resources" [3].

The word "identification' is not used anywhere in the RDF specs, I believe. 

> So if I read
>  <uri1> a :G-box .
> and then, by fetching <uri1>, I get a "200 Ok" and a Turtle
> representation (a g-text), then I would tend to consider that <uri1> the
> g-box "contains" the triples that I parse from the g-text.

That is the http-range-14 decision in a nutshell. However, what if you get a 303 redirect? What if you get a 404? What if the IRI has a hash in it somewhere? I am not meaning to imply that there are no answers to these questions, only that we need to provide them. 

> Note that I do not suggest that *any* g-box should be accessible through
> HTTP. I can think of hidden g-boxes, imaginary g-boxes...

Quite. I agree.


>  pa
> [1]
> [2]
> [3]

IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile

Received on Monday, 18 April 2011 17:58:59 UTC