Re: A modest attempt to re-open ISSUE-20

On 03/14/2013 12:49 PM, James Leigh wrote:
> On Thu, 2013-03-14 at 17:11 +0100, Pierre-Antoine Champin wrote:
>> James,
>>
>>
>> On Thu, Mar 14, 2013 at 3:07 PM, James Leigh <james@3roundstones.com>
>> wrote:
>>          On Thu, 2013-03-14 at 10:03 +0100, Henry Story wrote:
>>          >
>>          > The relevant section of the Turtle spec is section 6.3
>>          > http://www.w3.org/TR/turtle/#relative-iri
>>          >
>>          >
>>          > Notice that you cannot use NTriples to do what you want
>>          either, since
>>          > you cannot
>>          > know when creating a resource what the URI of the created
>>          resource is
>>          > going to be.
>>          > You would need to only use blank nodes in the graph
>>          produced, which
>>          > would be
>>          > to say the least very awkward.
>>
>>
>>
>>          In Callimachus, when a resource is created there is no funny
>>          base-uri
>>          going on (we use the RDF model). Although we use sparql-update
>>          format,
>>          any RDF format would work equally well for us including
>>          NTriples.
>>
>>          Lets calls the to-be-created resource URI the entity URI. All
>>          the
>>          triples that are to-be-inserted to the store must either have
>>          a blank
>>          node subject, have a subject URI of the entity URI, or have a
>>          subject
>>          URI of the entity URI plus a fragment identifier. This makes
>>          it very
>>          clear what the entity URI is.
>>
>>
>> But it prevents you from POSTing an *incoming* graph to the entity
>> resource...
>>
>> What if I want to create a person and state that "I know this
>> person" ?
>>
>
> In Callimachus, you would use two requests. The first one (POST) to
> create this new person resource and a second request (PATCH) to update
> another person with the statement "I know them". It's fairly easy and
> makes sense when you treat triples as belonging to their subject
> resource.

I agree with this approach. For me, this is Linked Data modeling done
right, and I would really like this group to generalize the notion of
"treat[ing] triples as belonging to their subject resource".

IMO, that should be the pattern that we'd encourage, along with the
use of hash-URIs.

Alexandre.

>
> Graphs are created/manipulated with a different set of requests and are
> handled separately.



>
> Regards,
> James
>>
>
>
>
>

Received on Thursday, 14 March 2013 17:22:37 UTC