Re: proposed second reply to Jeremy's comment about named graphs

On Sep 11, 2013, at 8:00 PM, Sandro Hawke wrote:

> On 09/11/2013 10:41 PM, Sandro Hawke wrote:
>> On 09/11/2013 08:43 PM, Pat Hayes wrote:
>>> I like this. Is it a good idea to also refer to the notes that Sandro and Pierre-Antoine are supposed to be writing? Just to show we havnt stopped worrying about it, you understand.
>> 
>> There's seems to be some lack of community memory on this.  I already gave Jeremy the formal reply to Jeremy's rdf:Graph comment, in which I explained about those two notes, etc [1].   He said he still wasn't happy [2].    I asked for more details [3], and he gave test cases [4] and proposed text [5].   I think we need to respond to *those* not to his earlier comments.
>> 
>> [1] http://lists.w3.org/Archives/Public/public-rdf-comments/2013Aug/0050.html
>> [2] http://lists.w3.org/Archives/Public/public-rdf-comments/2013Sep/0005.html
>> [3] http://lists.w3.org/Archives/Public/public-rdf-comments/2013Sep/0007.html
>> [4] http://lists.w3.org/Archives/Public/public-rdf-comments/2013Sep/0010.html
>> [5] http://lists.w3.org/Archives/Public/public-rdf-comments/2013Sep/0017.html
>> 
>> I haven't yet had a chance to read and think about [4] and [5].
>> 
> 
> Okay, I looked at them.    I don't see how one can accept both his proposed definition of rdf:Graph and his proposed outcome of TC1 (in [4]).

The original paper describing named graphs defined a named graph to be the pair <name, graph>, and gave it the semantics that the name denotes this pair. This would give Jeremy's response to TC1. I do not know if he is thinking of this, but it would be a consistent position to take. 

>    I think he's proposing that when rdf:Graph is used, the graph name denotes the graph,

the named graph, consisting of the pair <name, graph>. So <g1, {:a :b :c }> is a different named graph than <g2,  {:a :b :c }>

> but surely in that case the TC1 entailment must hold -- and he doesn't want it to -- because his mental model is really gboxes.

This name-graph pair device was an early small step in the direction of what we call gboxes: it allowed for there to be several distinct copies of a graph and the name to refer to one of them without referring to the others. But of course it did not provide for labile graph sources or things with state.

> 
> I guess I'll ask him.

Good idea.

BTW, I cannot understand his second proposed wording in [5]. " If a resource in this class is identified by an IRI, and that IRI is used to name a graph in a dataset, then within that dataset the resource SHOULD correspond to the named graph."

What does "correspond to" mean here? What does "identified" mean, exactly? 

I would vote NO on both wordings as they stand, as neither of them are precise enough to specify any semantics with any clarity. If any semantic (including referential) conditions are going to be imposed upon datasets, then they should be stated exactly in the Semantics document, using semantic terminology. However, this said, I agree with Peter that we have now decided that no such semantics is going to get defined by this WG.

Pat

> 
>        - s
> 
>>       -- Sandro
>> 
>> 
>> 
>>> Pat
>>> 
>>> On Sep 11, 2013, at 1:28 PM, Peter Patel-Schneider wrote:
>>> 
>>>> Dear Jeremy:
>>>> 
>>>> This is a second official response to your comment about named graphs in
>>>> http://lists.w3.org/Archives/Public/public-rdf-comments/2013Jul/0021.html and
>>>> http://lists.w3.org/Archives/Public/public-rdf-comments/2013Sep/0005.html 
>>>> 
>>>> 
>>>> The RDF Working Group believes that there are several ways in which RDF
>>>> graphs and datasets are and will be used.  These include ways that fit into
>>>> your use cases, where the graph names denote the graph they name or some
>>>> other formal graph-related construct and where you would indeed say
>>>> something like
>>>> 
>>>> jjc:graph {
>>>>   jjc:graph dc:creator "Jeremy J. Carroll" .
>>>> }
>>>> 
>>>> However, there are also ways that do not fit into your use cases, for
>>>> example where the graph names are IRIs that denote some other entity, such
>>>> as
>>>> 
>>>> jjc:jjc {
>>>>   jjc:jjc rdf:type foaf:Person .
>>>>   jjc:jjc foaf:lastName "Carroll" .
>>>>   jjc:jjc foaf:knows jjc:pfps .
>>>> }
>>>> 
>>>> If the RDF semantics required that all graph names denote graph-related
>>>> constructs this would interfere with these other use cases. Therefore the RDF
>>>> Working Group decided to not so require.
>>>> 
>>>> Further the RDF Working Group was unable to agree on even a weak theory of
>>>> named RDF graphs, such as one conditioned on explicit typing. Even the
>>>> nature of what graph names might denote was problematic: does the name of an
>>>> RDF graph denote the graph itself, does it denote some other construct that
>>>> is related to the graph, or does it even denote the semantic meaning of the
>>>> graph?
>>>> 
>>>> Therefore the working group has produced a very minimal specification for
>>>> RDF datasets and named graphs that does not depend on denotation.
>>>> 
>>>> This approach produces maximally compatability, but does not produce
>>>> inferences that might be desirable in some use cases.  If you do want
>>>> certain inferences to be part of your approach, such as the first example
>>>> above entailing
>>>>   jjc:graph rdf:type jjc:Graph.
>>>> you can define and implement a particular RDF entailment regime that
>>>> sanctions these inferences.
>>>> 
>>>> The RDF Working Group believes that this minimal approach will allow
>>>> different approaches to named graphs to coexist some allowing what you want
>>>> and others incompatible with what you want.  The flourishing approaches can
>>>> then be considered for standardization at a later time.
>>>> 
>>>> 
>>> ------------------------------------------------------------
>>> IHMC                                     (850)434 8903 home
>>> 40 South Alcaniz St.            (850)202 4416   office
>>> Pensacola                            (850)202 4440   fax
>>> FL 32502                              (850)291 0667   mobile (preferred)
>>> phayes@ihmc.us       http://www.ihmc.us/users/phayes
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> 
>> 
>> 
>> 
> 
> 
> 

------------------------------------------------------------
IHMC                                     (850)434 8903 home
40 South Alcaniz St.            (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile (preferred)
phayes@ihmc.us       http://www.ihmc.us/users/phayes

Received on Thursday, 12 September 2013 06:49:58 UTC