W3C home > Mailing lists > Public > public-rdf-comments@w3.org > May 2013

Re: Turtle data-type

From: Sandro Hawke <sandro@w3.org>
Date: Wed, 29 May 2013 08:11:24 -0400
Message-ID: <51A5F06C.1040004@w3.org>
To: Henry Story <henry.story@bblfish.net>
CC: Peter Ansell <ansell.peter@gmail.com>, Kingsley Idehen <kidehen@openlinksw.com>, public-rdf-comments@w3.org
I agree that referential opacity is not provided by RDF reification and 
that it is sometimes important. But I think (1) datasets provide it, and 
(2) as you suggest it can be provided with a datatype.  Since anyone can 
define a datatype, there's no need for RDF-WG to do so.  Maybe once one 
has been prototyped and used effectively, it could become a standard.

Also, procedurally, we've decided to use datasets as the way we handle 
multiple graphs.   Specifically, what you're talking about was I think 
part of what we called "Graph Literals".

    2012/05/23-rdf-wg RESOLVED:  Close ISSUE-5 ("Should we define Graph
    Literal datatypes?"), saying No, we should not.

So unless you have new information, it's not really something for the WG 
to talk about.  A few other relevant resolutions"

    2011/04/14-rdf-wg RESOLVED:  Named Graphs in SPARQL associate IRIs
    and graphs *but* they do not necessarily "name" graphs in the strict
    model-theoretic sense. A SPARQL Dataset does not establish graphs as
    referents of IRIs (relevant to ISSUE-30)

    2012/05/23-rdf-wg RESOLVED:  Close ISSUE-30 ("How does SPARQL's
    notion of RDF dataset relate our notion of multiple graphs?"),
    saying we will use SPARQL's notion of RDF dataset as much of the
    foundation of our handling of multiple graphs.

    2012/05/23-rdf-wg RESOLVED:  Close ISSUE-33 ("Do we provide a way to
    refer sub-graphs and/or individual triples?"), with the
    understanding that datasets can be used to refer to sub-graphs and
    individual triples. This does NOT rule out sharing blank nodes
    between named graphs.

    2012/10/03-rdf-wg RESOLVED:  This Working Group will not provide a
    Formal Semantics for RDF Datasets or for our Dataset Syntax (eg
    trig). The WG may publish some information about dataset semantics
    in WG NOTES.

    2012/10/30-rdf-wg RESOLVED:  The RDF WG intends to produce a NOTE on
    the semantics of datasets.

    2012/10/30-rdf-wg RESOLVED:  The WG intends to produce a NOTE:
    Practical Use Cases of RDF Datasets.  This Note may include
    information from ISSUEs 32, 35 and 38.

    2013/01/23-rdf-wg RESOLVED:  Close ISSUE-25 by saying that this WG
    will not deprecate reification of statements.  We will note
    informatively in the RDF Schema spec that named graphs and RDF
    datasets are another mechanism to accomplish the same goals.

These are from http://www.w3.org/2011/rdf-wg/resolutions.txt which is 
kind of lame (sorry), but better than nothing.  Kingsley, please see 
http://www.w3.org/2011/rdf-wg/meetings which is also kind of lame but 
better than nothing.

As an aside, I continue to see little benefit to distinguishing between 
datatypes and (functional, string-valued) properties. It's like 
distinguishing between a salad fork and a dinner fork.    But salad 
forks seem to be a thing people like, so I guess it's okay.

        -- Sandro


On 05/29/2013 02:31 AM, Henry Story wrote:
>
> On 29 May 2013, at 07:38, Peter Ansell <ansell.peter@gmail.com 
> <mailto:ansell.peter@gmail.com>> wrote:
>
>> On 29 May 2013 15:29, Henry Story <henry.story@bblfish.net 
>> <mailto:henry.story@bblfish.net>> wrote:
>>
>>
>>     On 28 May 2013, at 21:41, Kingsley Idehen <kidehen@openlinksw.com
>>     <mailto:kidehen@openlinksw.com>> wrote:
>>
>>     > On 5/20/13 11:39 AM, Henry Story wrote:
>>     >> In Turtle there is no way of specifying a graph (other than
>>     through reificiation which
>>     >> is understood to be broken).
>>     >
>>     > Why is reification assumed to be broken? I think the perception
>>     is that its cumbersome due to data bloat. Like blank nodes, this
>>     is a feature of RDF that's often misunderstood and then in the
>>     process maligned.
>>     >
>>     > Reification is the powerful mechanism for granular descriptions
>>     of triples (statements) without exiting existing RDF semantics.
>>
>>     It does not do quotation the way we need it.
>>
>>     I want to express the following:
>>
>>     Laura Lane believes that Superman is a Hero .
>>     Superman is Clark Kent .
>>
>>     With the suggested notation .
>>
>>     lane:Laura believes """@prefix ...
>>        super:man a Hero .
>>     """^^lang:Turtle .
>>     super:man owl:sameAs kent:Clark .
>>
>>     With reification
>>
>>     lane:Laura believes _:b .
>>     _:b rdf:subject super:man ;
>>         rdf:relation rdf:type ;
>>         rdf:object Hero .
>>     super:man owl:sameAs kent:Clark .
>>
>>     Now since owl:sameAs allows substitution of identicals salva
>>     veritate,
>>     it follows from the reification example but not from the
>>     quotation example
>>     that Laura Lane believes that Clark Kent is a Hero.
>>
>>
>> IMO, Laura Lane *would* believe that Clark Kent is a Hero if she were 
>> privileged to know the information inside of *either* the quotation 
>> or the reification.
>
> To be precise if:
>
>   lane:Laura believes ".... super:man owl:sameAs kent:Clark 
> "^^lang:Turtle
>
> then you could come to the conclusion that she should believe that 
> Clark Kent is a Hero.
> But that is something that may or may not be true.
>
>
>> Are you trying to create an alternative for reification for use in 
>> RDF that is specifically not compatible with OWL to avoid conflicts 
>> with the OWL semantics?
>
> Not at all.
>
> The Laura Lane example is just a modernization of examples that stem 
> from Frege, and that
> have been analysed in philosophy of language and logic since the 
> beginning of the subject.
>
> In the semantic web this has clearly been accepted with SPARQL graphs, 
> NQuads,
> N3, .... I am just pointing out that this has been available if hidden 
> in RDF from the
> beginning.
>
>>     You can argue that is not broken because it is not the same as
>>     quotation
>>     and it just does something else, and that would be fine.
>>
>>     But we need the quotation mechanism, for many use case. It is not
>>     bringing any new
>>     semantics into RDF in any case: it's just asking for a notation to
>>     express what you can already express. We'd just like a standard
>>     notation
>>     for something as important as this. defining lang:Turtle as a
>>     datatype,
>>     with a well known w3c url.
>>
>>
>> So if there are no new semantics, does that imply that the quotation 
>> has no effect unless the interpreter manually extracts the Turtle 
>> from the Literal and interprets it outside of the RDF graph derived 
>> from the document?
>
> Indeed. You need a further non-automatic step to extract content from 
> a quoted context.
> And more to the point you need a further step to add something to a 
> quoted context.
> If you want to come to an idea of what Laura Lane believes you need to 
> add a statement to
> her belief context. It is not because something is true that everyone 
> believes it to be the case.
>
>>
>> Are you able to use the existing well-known W3C URL for Turtle, 
>> http://www.w3.org/ns/formats/Turtle , and others from the list at 
>> http://www.w3.org/ns/formats/ in a similar way?
>
> $ cwm http://www.w3.org/ns/formats/Turtle
>      @prefix : <http://xmlns.com/foaf/0.1/> .
>     @prefix dcterms: <http://purl.org/dc/terms/> .
>     @prefix foa: <http://www.ivan-herman.net/foaf#> .
>     @prefix formats: <./> .
>     @prefix rdfs: </2000/01/rdf-schema#> .
>     foa:me     a :Person;
>          rdfs:seeAlso <http://www.ivan-herman.net/foaf>;
>          :mbox <mailto:ivan@w3.org>;
>          :name "Ivan Herman";
>          :title "Semantic Web Activity Lead";
>          :workplaceHomepage <http://www.w3.org> .
>     formats:Turtle     a formats:Format;
>          dcterms:created "2010-05-03";
>          dcterms:creator foa:me;
>          dcterms:description "Unique identifier for the RDF 
> Serialization in Turtle";
>          dcterms:modified "2013-03-25";
>          rdfs:comment "Turtle is defined by the “Turtle - Terse RDF 
> Triple Language” W3C Candidate Recommendation.";
>          rdfs:isDefinedBy </TR/turtle/>;
>          rdfs:seeAlso </TR/turtle/#sec-mediaReg>;
>          formats:media_type "text/turtle";
>          formats:preferred_suffix ".ttl" .
>
> I am not entirely sure. I could not dereference 
> http://www.w3.org/ns/formats/
> to see if it is a datatype definition.
>
> But if it were a Datatype definition then one could have a datatype 
> for each of the
> serialisations. It would make sense to do so. I limited myself to 
> Turtle here as
> I have a use case for it.
>
>
>
>>
>> Peter
>
> Social Web Architect
> http://bblfish.net/
>
Received on Wednesday, 29 May 2013 12:11:42 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:29:56 UTC