Re: different Semantics proposals (Re: Agenda for 19 Sep 2012)

After reading and following all this debate, I agree with Peter. 

First, I do not see any sign of any form of consensus on the relationship between a default graph and the named graphs in a dataset. We have seen at least three different ideas (default=union; default=metadata; default=accepted distillation) in our own discussions come close to being 'standardized'. So any dataset semantics which imposes such a constraint (between default graph and named graphs) should be treated with at least suspicion, and probably be rejected. 

Second, I do not see why we need to give a semantics for datasets. The graphs in a dataset already have a semantics, and we can describe entailment relations between them and users can specify those relations between graphs in one dataset and graphs in a different dataset, in combinations which suite their purposes, and nothing is in any way ambiguous or underspecified. The one exception to this would be to formalize the useful notion of naming, ie the relationship between a URI N and a graph G which is supposed to be indicated by the presence of the pairing <N, G> in a dataset. But the "minimal semantics" does not do this in a satisfactory way: it does not determine a single graph named by the URI, and it does not relate this <N, G> relationship to the identification relationship between URIs and resources corresponding to the HTTP protocols, or to existing use of names in RDF. Since the RDF semantics makes no reference to the new G-EXT relationships, there is no way to use the N to refer to the G in any RDF metadata. The "naming" provided by the minimal semantics is therefore of no practical use, and is likely to lead to confusion.

On balance, I think that to impose even this minimal semantics as a normative standard would be pointless and potentially conterproductive. It has no obvious purpose, it achieves nothing useful, it will cause confusion and it might cause harm. It would be better to do nothing.

Pat


On Sep 18, 2012, at 11:47 AM, Peter F. Patel-Schneider wrote:

> 
> On 09/18/2012 12:26 PM, Antoine Zimmermann wrote:
> [...]
> 
>> At the moment, I'm convinced there are many private ways of doing things, but I don't think there are many exchange of datasets between systems, such that the need for a sanctioned semantics have been limited. But we see a number of use cases that are either happening right now, or that are going to appear inevitably, which requires a common understanding of what dataset semantics is.
>> 
>> Of course, we want the least common denominator such that as many current practices as possible are conforming.
>> 
>> But we may also make bold decisions that prevent what we consider bad practices, even if it breaks some implementations.
>> 
>> 
>> -AZ
>> 
> 
> I agree in general with this reasoning, but come up with a different end point.
> 
> If systems are storing RDF datasets, and exchanging them (even if not globally), and have a very limited (or even null) theory of the interaction between the default graph and the named graphs, then we should be very careful in breaking this form of interaction.  This would be another rationale for not having information bleed from the default graph into the named graphs.
> 
> Any bold decisions then have to have convincing rationale.   There has to be something very useful that comes from the boldness and there can't be much broken that isn't totally unredeemable.
> 
> peter
> 
> 
> 
> 

------------------------------------------------------------
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
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Tuesday, 18 September 2012 17:42:18 UTC