Re: What is Oracle's objection to the use of Turtle as R2RML syntax?

On 12/13/11 6:40 PM, Souripriya Das wrote:
> Kingsley,
>
> As a precedence, we can look at the OWL spec. OWL too was called a language, "Web Ontology Language". But, it just defined a vocabulary. It did not define any syntax. One can use any RDF syntax (RDF/XML, N-Triple, ...) for OWL.

Souri,

Yes, you could look at OWL as a DSL for expressing the semantic fidelity 
of relations via description logics. Now, if you look at OWL, as 
powerful and fundamentally useful as it actually is, what's happened to 
it over the last 12+ years? Nothing but confusion due to syntax level 
issues, all at the expense of its underlying syntax agnostic model.

Fundamentally, our goal should be to learn key lessons from the past 
re., devastating effects too many syntaxes for achieving a pragmatic 
goal such as declarative mappings between the RDBMS and RDF  realms re., 
R2RML.

Having Turtle as the base language for interoperability is inherently 
advantageous. Bring RDF/XML and other syntaxes into the mix and we have 
nothing but adoption inertia all over again.

Kingsley
>
> Thanks,
> - Souri.
>
> ----- Original Message -----
> From: kidehen@openlinksw.com
> To: public-rdb2rdf-wg@w3.org
> Sent: Tuesday, December 13, 2011 5:01:24 PM GMT -05:00 US/Canada Eastern
> Subject: Re: What is Oracle's objection to the use of Turtle as R2RML syntax?
>
> On 12/13/11 4:51 PM, Richard Cyganiak wrote:
>> Hi Ashok,
>>
>> On 13 Dec 2011, at 21:11, ashok malhotra wrote:
>>> At the Linked Data Workshop last week IBM showed some slides that used a RDF syntax called Trig.
>>> So, I looked up Trig http://www.wiwiss.fu-berlin.de/suhl/bizer/TriG/Spec/ and found that you were
>>> one of the authors.  So, Trig might be an alternate syntax for the mapping language, no?
>>> Folks seemed to like it.  It is not a standard but may become one.
>> I like that example! R2RML already allows the use of TriG syntax, no change required. TriG is a superset of Turtle. If you serialize an R2RML mapping graph as TriG, you get a Turtle file.
>>
>>> It is also possible that
>>> other RDF syntaxes will appear and one of them will become dominant.  If that happens,
>>> it would be good if users could  write R2RML in the new syntax.
>> It takes a while for a new syntax to be invented, be implemented, be evangelized, become popular, and become a W3C Recommendation. In the case of Turtle, 13 years. A future RDB2RDF WG can relax the Turtle requirement for a future R2RML version if demand for other syntaxes materializes. This is not something that needs to be considered for R2RML 1.0.
>>
>> [[
>> PROPOSAL: On ISSUE-57, let's just go with Turtle in order to get R2RML 1.0 out of the door.
> +1
>
> Kingsley
>> ]]
>>
>> Best,
>> Richard
>>
>>
>>
>>> All the best, Ashok
>>>
>>> On 12/13/2011 12:07 PM, Richard Cyganiak wrote:
>>>> Souri, what you say here is all correct (well, some nitpicks inline), but I still don't see the key question addressed: What makes one-w3c-RDF-syntax-MUST better in your eyes? All you do below is show that users and implementers will have to go through extra hoops if that proposal is accepted, so you're actually sort of making a case against it…
>>>>
>>>> On 13 Dec 2011, at 16:34, Souripriya Das wrote:
>>>>> There are currently two proposed options:
>>>>> 1) Turtle-syntax-MUST
>>>>> 2) one-w3c-RDF-syntax-MUST
>>>>>
>>>>> The Turtle-syntax-MUST option does not support any syntax other than Turtle for the R2RML mapping documents
>>>> (Right, R2RML *mapping documents* MUST be Turtle. But implementations MAY still support any syntax other than Turtle.)
>>>>
>>>>> while the one-w3c-RDF-syntax-MUST option only requires at least one of the (W3C Recommendation) RDF syntaxes be used for R2RML mapping documents
>>>>> (while, because of the Turtle-convertibility,  still allowing the test cases, tutorials, books, etc. to be written in Turtle).
>>>> But one-w3c-RDF-syntax-MUST would equally allow tutorials and books to be written in any other syntax, so we'll get books that teach R2RML using RDF/XML syntax, tutorials that teach R2RML using RDFa syntax, et cetera.
>>>>
>>>>> Again, here is the one-w3c-RDF-syntax-MUST proposal:
>>>>> [[
>>>>> An R2RML mapping document is any document that encodes an R2RML mapping graph and
>>>>> is written in any RDF syntax that is a W3C Recommendation and that can be converted to Turtle.
>>>>> ]]
>>>>> (By "Turtle" we mean the future Turtle W3C Recommendation)
>>>>>
>>>>> The following can be said about the effect of using the above proposal:
>>>>>
>>>>> ------------------
>>>>> Consider an R2RML mapping document written in RDF/XML syntax:
>>>>> - Is it a conforming R2RML mapping document?
>>>>> YES. (for Turtle-syntax-MUST: NO)
>>>> …and this is literally the *only* difference. I don't understand who benefits from this. Who is keen on calling an R2RML mapping graph serialized in RDF/XML an R2RML mapping document, and why?
>>>>
>>>>> - Why?
>>>>> Because one can convert the document to generate an equivalent document written in Turtle syntax.
>>>>>
>>>>> ---------------------
>>>>> Consider an R2RML mapping processor which ONLY accepts R2RML mapping documents written in RDF/XML syntax:
>>>>> - Is it a conforming R2RML mapping processor?
>>>>> YES. (for Turtle-syntax-MUST: YES)
>>>>> - Why?
>>>>> Because it accepts all conforming R2RML mapping documents, written in RDF/XML syntax.
>>>>>
>>>>> ---------------------
>>>>> Testing for conformance of an R2RML mapping processor that ONLY accepts RDF/XML documents:
>>>> Yes, it's possible, but it's inconvenient for implementers. What advantage balances this inconvenience?
>>>>
>>>>> For each test in "tests for conformance"
>>>>> 1) obtain the mapping documents (written in Turtle syntax)
>>>>> 2) Convert these Turtle-syntax mapping documents into RDF/XML syntax (assuming this is possible)
>>>>> 3) Process the converted documents
>>>>> 4) Run the corresponding SPARQL queries from "tests for conformance" and compare the results
>>>>> 5) If query results match, then this processor is indeed a conforming R2RML mapping processor
>>>>>
>>>>> ---------------------
>>>>> Sharing of an R2RML mapping document between two non-overlapping syntax accepting processors:
>>>> Yes, it's possible, but it's inconvenient for mapping authors and admins. What advantage balances this inconvenience?
>>>>
>>>> I'm not saying that there is no advantage. I'm just saying that I myself can't think of any, and that I can't recall anyone else mentioning any!
>>>>
>>>> So, which stakeholder (WG members, implementers [incl. Oracle], mapping authors) benefits from one-w3c-RDF-syntax-MUST, and how?
>>>>
>>>> Best,
>>>> Richard
>>>>
>>>>
>>>>> - MappingProcessor1 accepts ONLY RDF/XML and MappingProcessor2 accepts ONLY N-Triples
>>>>> - an R2RML mapping document (written in RDF/XML) and used at MappingProcessor1 is to be shared with MappingProcessor2
>>>>> - convert the mapping document into an equivalent N-Triples document (via Turtle, if direct conversion is not possible)
>>>>> - present the N-Triples R2RML mapping document to MappingProcessor2
>>>>>
>>>>> Thanks,
>>>>> - Souri.
>>>>>
>>>>> ----- Original Message -----
>>>>> From: richard@cyganiak.de
>>>>> To: souripriya.das@oracle.com
>>>>> Cc: public-rdb2rdf-wg@w3.org
>>>>> Sent: Wednesday, December 7, 2011 1:52:14 PM GMT -05:00 US/Canada Eastern
>>>>> Subject: Re: What is Oracle's objection to the use of Turtle as R2RML syntax?
>>>>>
>>>>> On 7 Dec 2011, at 18:07, Souripriya Das wrote:
>>>>>> Instead of the following definition of the mapping document:
>>>>>>
>>>>>> [[
>>>>>> An R2RML mapping document is any document written in the Turtle [TURTLE] RDF syntax that encodes an R2RML mapping graph.
>>>>>> ]]
>>>>>>
>>>>>> we propose the following:
>>>>>>
>>>>>> [[
>>>>>> An R2RML mapping document is any document that encodes an R2RML mapping graph and
>>>>>> is written in any RDF syntax that is a W3C Recommendation and can be converted to Turtle [2].
>>>>>> ]]
>>>>> Why is this better?
>>>>>
>>>>> Richard
>>>>>
>>
>


-- 

Regards,

Kingsley Idehen 
Founder&  CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen