W3C home > Mailing lists > Public > public-rdf-wg@w3.org > August 2013

Re: Comment on rdf concepts (RDF-ISSUE-140)

From: Andy Seaborne <andy@apache.org>
Date: Fri, 09 Aug 2013 17:32:56 +0100
Message-ID: <520519B8.6000705@apache.org>
To: Ivan Herman <ivan@w3.org>
CC: public-rdf-wg@w3.org
On 09/08/13 10:17, Ivan Herman wrote:
>
> On Aug 8, 2013, at 24:36 , Andy Seaborne <andy@apache.org> wrote:
>
>> On 07/08/13 21:02, Markus Lanthaler wrote:
>>> On Monday, August 05, 2013 6:29 PM, Ivan Herman wrote:
>>>> I have a question/comment on this:
>>>>
>>>> [[
>>>> 4.1 RDF Dataset Comparison
>>>>
>>>> Two RDF datasets (the RDF dataset D1 with default graph DG1 and named
>>>> graph NG1 and the RDF dataset D2 with default graph DG2 and named graph
>>>> NG2) are dataset-isomorphic if and only if:
>>>>
>>>> 	. DG1 and DG2 are graph-isomorphic;
>>>> 	. For each (n1,g1) in NG1, there exists (n2,g2) in NG2 such that
>>>> n1=n2 and g1 and g2 are graph-isomorphic;
>>>> 	. For each (n2,g2) in NG2, there exists (n1,g1) in NG1 such that
>>>> n1=n2 and g1 and g2 are graph-isomorphic.
>>>>
>>>> ]]
>>>>
>>>> A graph name can now be a blank node. Wouldn't it be appropriate to use
>>>> the 'M' mapping of section 3.6 for the graph names, too? Or are we
>>>> deliberately silent on this?
>>>
>>> Hmm... I'm not sure, it would probably make sense and be consistent with the
>>> rest. On the other hand, since the graph names are just labels it may not
>>> make much sense because you could argue that two blank node "labels" will
>>> never match as they are dataset-local.
>>
>> This is defining "dataset-isomorphic" (i.e. it's a data structure matter, not "equality") so the graph label should be included in the isomorphism-ness.
>>
>
> Actually... we did say that two graphs in the same dataset share bnodes. On the other hand, if I look at the definition above, I am not sure it accounts for this fact. The isomorphism of the constituent graphs is reused separately from one another. If I have
>
> DS1:
> <A> { a b _:b1 }
> <B> { p q _:b1 }
>
> DS2:
> <A> {a b _:x1 }
> <B> {p q _:y1 }
>
> then DS1 and DS2 are dataset-isomorphic according to this definition. Is this what we want?
>
> I would think that dataset isomorphism should require that I have to use _one_ bijection 'M', as described in 3.6, defined on the union of all the graph nodes plus the union of the graph labels. Which would automatically take care of my original issue, too.
>
> Do I miss something here?
>
> Ivan

I'm expecting the definition of dataset-isomorphism to be a single space 
of bnodes, to reflect bNode sharing across graphs and use in graph 
names.  i.e. one M function.

	Andy

>
>
>
>
>
>
>> 	Andy
>>
>> PS Testing needs this.
>>
>>> Anyway, I've opened ISSUE-140 to keep track of this:
>>>    https://www.w3.org/2011/rdf-wg/track/issues/140
>>>
>>>
>>> Cheers,
>>> Markus
>>>
>>>
>>> --
>>> Markus Lanthaler
>>> @markuslanthaler
>>>
>>>
>>>
>>>
>>>
>>
>>
>
>
> ----
> Ivan Herman, W3C
> Home: http://www.w3.org/People/Ivan/
> mobile: +31-641044153
> FOAF: http://www.ivan-herman.net/foaf.rdf
>
>
>
>
>
Received on Friday, 9 August 2013 16:33:24 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:04:31 UTC