Re: Union or not union for the default graph...

Hi all,

Perhaps we should provide a standard way for an RDF system to advertise how the default graph. That would be sufficient to plug the hole Ivan sees without forking the semantics, wouldn't it?

Linked Data systems already advertise a lot via VoID. This wouldn't require much of an extension and needn't even be normative. 

Regards,
Dave


On Apr 12, 2012, at 9:36, Andy Seaborne <andy.seaborne@epimorphics.com> wrote:

> 
> 
> On 12/04/12 14:19, Ivan Herman wrote:
>> 
>> On Apr 12, 2012, at 15:12 , Andy Seaborne wrote:
>> 
>>>>> Should that be reflected in the TriG file? Actually, maybe
>>>>> not.
>>> 
>>> I confess I don't see the need for this in the semantics.
>>> 
>>> Whether the default graph is treated as the union of the named
>>> graphs is a decision made by the application (see William's point),
>>> in which case the default graph is just some graph it has (locally)
>>> created.  It will do that based on whether it thinks the incoming
>>> data is trustworthy and appropriate to the task.
>>> 
>>> It might be useful to (non-normatively) discussion default union
>>> graphs but it is just a union graph like any other.
>>> 
>>> It might be useful to be able to have syntax to say "for this TriG
>>> file make the default graph the union of named graph" but that is a
>>> syntactic shortcut to avoid having to repeat the data, no semantics
>>> implications.
>>> 
>> 
>> What seems to be the case is that (a) there are use cases where the
>> union semantics is needed
> 
> Hmm - I see use cases which can be addressed by the application deciding to use the default union graph.  I'm not sure that I see any where it is needed as part of the semweb web-wide architecture, i.e. the publishing entity making the declaration and expecting it to be the only way it's used.
> 
>> and (b) this may not be the only semantics,
> 
> :-) so many to choose from
> 
>> so there is a need for a switch. Now we can put that switch in the
>> TriG file via some syntax (which is what you seem to propose)or an
>> out-of-band switch that is used to control the implementation; that
>> what I was proposing. I would not _really_ care as long as the choice
>> is clear (and defining the semantics is a way of making it really
>> clear) and there is a way to make the choice, although Steve's recent
>> mail seems to indicate that for his system at least the 'out-of-band'
>> approach may be more appropriate.
>> 
>> Cheers
>> 
>> I.
> 
>    Andy
> 
>> 
>>> Andy
>>> 
>>> On 12/04/12 13:35, Ivan Herman wrote:
>>>> I have sent this to the wrong mailing list yesterday evening;
>>>> Andy pinged me that I did that:-( so here is the mail again.
>>> 
>>> and only because it didn't end up in the right folder.
>>> 
>>>> 
>>>> Since then I have also added the text to the Wiki page.
>>>> 
>>>> Ivan
>>>> 
>>>> Begin forwarded message:
>>>> 
>>>>> From: Ivan Herman<ivan@w3.org> Subject: union or not union for
>>>>> the default graph... Date: April 11, 2012 21:21:41 GMT+02:00
>>>>> To: Sandro Hawke<sandro@w3.org>, Andy
>>>>> Seaborne<andy.seaborne@epimorphics.com>, Pat
>>>>> Hayes<phayes@ihmc.us> Cc: W3C Public RDF Mailing
>>>>> list<public-rdf@w3.org>
>>>>> 
>>>>> Just some thoughts late at night for me... I just realized
>>>>> something.
>>>>> 
>>>>> I think we may have to acknowledge that what I would call the
>>>>> "quoting" semantics (which is the one on the wiki page) and a
>>>>> "union" semantics may both be necessary, depending on the
>>>>> application. Tom's frbr example needed the union semantics, ie,
>>>>> where the default graph includes the union of all constituent
>>>>> graphs in the datase for the interpretationt, whereas the
>>>>> "quote" semantics is clearly used in other settings. We know
>>>>> that SPARQL engines exist that operate with one, others with
>>>>> the other.
>>>>> 
>>>>> Defining the two semantics formally is obvious. The "quote"
>>>>> semantics is the one on the wiki page; the "union" semantics
>>>>> differs only in term of the definion of the vocabulary which
>>>>> would have to include the union of all graphs, too. Otherwise
>>>>> it is the same.
>>>>> 
>>>>> Should that be reflected in the TriG file? Actually, maybe not.
>>>>> It depends on the application and the underlying system which
>>>>> semantics is used. The same dataset can be interpreted this way
>>>>> or that way. We already have such a situation: the same OWL RL
>>>>> data can be interpreted through the Direct Semantics and the
>>>>> RDF Compatible Semantics. There are subtle differences between
>>>>> the two and it is up to the application to choose. The
>>>>> difference between the "quoting" and the "union" semantics is
>>>>> much less subtle, but the situation is similar. We may also
>>>>> have engines that do RDF semantics, others do D-entailment, or
>>>>> RDFS entailment...
>>>>> 
>>>>> Bottom line: we may define two different dataset semantics (ie,
>>>>> two different interpretations) and... leave it that.
>>>>> 
>>>>> Thoughts?
>>>>> 
>>>>> Ivan
>>>>> 
>>>>> --- Ivan Herman Tel:+31 641044153 http://www.ivan-herman.net
>>>>> 
>>>>> (Written on mobile, sorry for brevity and misspellings...)
>>>>> 
>>>>> 
>>>> 
>>>> 
>>>> ---- Ivan Herman, W3C Semantic Web Activity Lead Home:
>>>> http://www.w3.org/People/Ivan/ mobile: +31-641044153 FOAF:
>>>> http://www.ivan-herman.net/foaf.rdf
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>> 
>> 
>> ---- Ivan Herman, W3C Semantic Web Activity Lead Home:
>> http://www.w3.org/People/Ivan/ mobile: +31-641044153 FOAF:
>> http://www.ivan-herman.net/foaf.rdf
>> 
>> 
>> 
>> 
>> 
>

Received on Thursday, 12 April 2012 20:32:07 UTC