W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > July to September 2011

Re: Update protocol and dataset description

From: Andy Seaborne <andy.seaborne@epimorphics.com>
Date: Mon, 08 Aug 2011 21:23:30 +0100
Message-ID: <4E4045C2.7070603@epimorphics.com>
To: public-rdf-dawg@w3.org


On 08/08/11 20:21, Lee Feigenbaum wrote:
> On 8/8/2011 2:57 PM, Andy Seaborne wrote:
>> The question is what happens within one request when there are two
>> operations.
>>
>> In your example, it's really a subset of the graph store (as I
>> understand it).
>>
>> In my example, the dataset is unchanging. No updates from the first
>> operation are visible to in the second operation (neither changes of a
>> graphs triples, not the set of named graphs).
>>
>> It's different from query because of multiple operations in one request.
>
> Maybe I need to see a more complete example. Are you saying that you'd
> pass something like http://example/web in the protocol parameter and
> then do:
>
> INSERT DATA { GRAPH <http://example/web> { :a :b :c } }
> ;
> INSERT {
> ...
> } WHERE {
> GRAPH <http://example/web> { :a :b ?c }
> }
>
> and not expect the 2nd operation to see the triple from the first?

There are two cases here; when http://example/web is in the protocol 
parameters and where http://example/web is not in the the protocol 
parameters and then the request is:

INSERT DATA { GRAPH <http://example/web> { :a :b :c } }

INSERT {
...
} WHERE {
   GRAPH <http://example/web> { :a :b ?c }
}

or GRAPH ?g { ... }

if the dataset is fixed by reading the web, there are two 
"http://example/web" in circulation.

(Note: I'm not suggesting one design over another - as the test cases 
attempted, I'm exploring the possibilites of the design space.  We have 
had little-to-no discussion of this so as far as I'm concerned I haven't 
formed a solidly based design-space understanding yet. Anzo is one case 
but it is not the only one.)

	Andy


>
> Lee
>
>>
>> Andy
>>
>> On 08/08/11 16:27, Lee Feigenbaum wrote:
>>> This seems like it would be well-handled by the semantics I'm intending
>>> here -- the dataset is given in the protocol request, which is basically
>>> as if it were given in USING/USING NAMED, at which points it's up to the
>>> implementation as to how to get at the graphs (via Web dereference or
>>> via local lookup or via something else entirely).
>>>
>>> Lee
>>>
>>> On 8/8/2011 10:16 AM, Andy Seaborne wrote:
>>>> I'll add a use case then:
>>>>
>>>> The ability to pull in data from the web and use as dataset in an
>>>> update
>>>> request.
>>>>
>>>> Here we have a temporary datasets for the update request, which creates
>>>> and extracts the data from within the request dataset and adds/deletes
>>>> it from the graph store.
>>>>
>>>> The dataset graphs do not exist in the local graph store before, during
>>>> or after the request. The entire WHERE part has been retargetted.
>>>>
>>>> Andy
>>>>
>>>>
>>>
>>
>>
>
Received on Monday, 8 August 2011 20:24:01 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:46 GMT