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

Re: Comments on Last-Call Working Draft of RDF 1.1 Semantics

From: Antoine Zimmermann <antoine.zimmermann@emse.fr>
Date: Wed, 11 Dec 2013 16:26:01 +0100
Message-ID: <52A88409.8070303@emse.fr>
To: Sandro Hawke <sandro@w3.org>
CC: Pat Hayes <phayes@ihmc.us>, Michael Schneider <schneid@fzi.de>, "public-rdf-comments@w3.org Comments" <public-rdf-comments@w3.org>

Le 11/12/2013 14:37, Sandro Hawke a écrit :
> On 12/11/2013 07:52 AM, Antoine Zimmermann wrote:
>> Le 10/12/2013 09:22, Pat Hayes a écrit :
>>> On Dec 9, 2013, at 9:12 PM, Pat Hayes <phayes@ihmc.us> wrote:
>>>> On Dec 9, 2013, at 4:15 PM, Michael Schneider <schneid@fzi.de>
>>>> wrote:
>>> ....
>>>>> There will be no further involvement into the discussion from my
>>>>> side, except for answering concrete requests, e.g. for
>>>>> clarification, etc. It's now up to the WG to make a decision.
>>>>> What I can say is that if this change makes it into PR, I'm going
>>>>> to formally object, and my basic line of argumentation should be
>>>>> clear by now.
>>>> It is not clear to me. As the change does not affect any
>>>> entailments and does not alter the actual interpretation structures
>>>> being described, and as the documents now (in reponse to your
>>>> original comment) exactly define the 2004 notion of datatype map so
>>>> as to provide backwards compatibility with earlier specifications
>>>> which use the concept, I do not see what the basis is for your
>>>> objection, other than that you prefer the older style of
>>>> exposition.
>>> Let me slightly modify my stance here. Since writing this I have
>>> taken a careful look at the 2010 OWL 2 RDF specification you edited,
>>> and I see that you there make very detailed reference to much of the
>>> formal machinery defined in the 2004 RDF Semantics document,
>>> repeating and then modifying or extending many of the exact
>>> definitions, including those of datatype maps. In order to refer to
>>> the RDF 1.1 Semantics rather than the 2004 Semantics in the same way
>>> would require, I now see, extensive editorial changes to the 2010 OWL
>>> 2 RDF document, since many of those definitions have now been
>>> modified or in some cases removed from the RDF 1.1 Semantics
>>> altogether. However, the particular issue we are discussing here
>>> seems to be only one of these changes, and not one that affects any
>>> entailments in either RDF or OWL 2. It would be straighforward, if
>>> tedious, to re-word the relevant sections (chiefly 4.1 and 4.2, with
>>> minor wording changes in 3.3 and 3.4) so that they used the concepts
>>> of D-interpretation and recognized IRIs defined in the RDF 1.1
>>> Semantics document, rather than referring to the datatype map
>>> construct used in the 2004 RDF specifications, while still describing
>>> exactly the same extended semantic structures related to RDF in the
>>> same way. (Other differences between the 2004 and 2013 RDF semantics
>>> descriptions seem to have more far-reaching editorial consequences;
>>> for example, that intepretations in RDF 1.1 are not defined with
>>> respect to a vocabulary. Others require changes to actual
>>> entailments, in particular the fact that ill-formed literals produce
>>> inconsistencies in RDF without using the RDFS extension.)
>>> Nevertheless, in order to make the connection to the 2004 definitions
>>> more explicit, let me suggest that we change our wording by adding to
>>> the end of the first paragraph of section 7 so that it reads as
>>> follows:
>>> -----
>>> Datatypes are identified by IRIs. Interpretations will vary according
>>> to which IRIs they recognize as denoting datatypes. We describe this
>>> using a parameter D on simple interpretations. where D is the set of
>>> recognized datatype IRIs. We assume that a recognized IRI identifies
>>> a unique datatype wherever it occurs, and the semantics requires that
>>> it refers to this identified datatype. The exact mechanism by which
>>> an IRI identifies a datatype IRI is considered to be external to the
>>> semantics, but we may describe this identification as a mapping from
>>> recognized IRIs to datatypes, called a <def>datatype map</def>.
>>> and then modify the change note as follows:
>>> <Change note> In the 2004 Semantics specification, D was defined to
>>> be a datatype map rather than a set of IRIs.  This  permitted
>>> "non-standard" datatype maps which map IRIs to datatypes they do not
>>> identify (such as one that maps the IRI
>>> 'http://www.w3.org/2001/XMLSchema#decimal' to the datatype identified
>>> by http://www.w3.org/2001/XMLSchema#gYearMonth ). Semantic extensions
>>> based on such non-standard mappings are not sanctioned by this
>>> specification.
>> Leaving aside the rest of the proposal for the moment, this change
>> note must not appear in the spec. Having datatype maps or sets does
>> not make any difference wrt this issue, and in any case,
>> "pathological" datatype maps were ruled out of the definition long
>> before an Editor's draft of Semantics even existed.
> Are you saying they were ruled out in the 2004 spec?   I'm looking at
> http://www.w3.org/TR/rdf-mt/#DTYPEINTERP and I don't see something like
> that.

No, the pathological datatype maps were ruled out of the definition of 
datatype map in RDF 1.1 Concepts Working drafts and editors draft 
spanning from May 2012 to May 2013. So the exclusion of "bad" datatype 
map is not something that is brought by the new design introduced by Pat 
in February 2013 in the first RDF 1.1 Semantics ED.

That is, claiming that the new design solves this issue is a fallacy. It 
was already solved before.

>> See RDF 1.1 Concepts WD in June 2012, revised in January 2013, and the
>> ED still contained the definition until May 2013.
> (I'm confused how those might be relevant.)
>> It has been your most repeated argument against datatype map, but it
>> is falacious.
> The argument I hear Pat making against datatype maps is that they are an
> unnecessary complication, one more variable than we actually need.
> That's the argument that resonates with me, although I don't claim to
> understand all facets of this matter.

Simplication is indeed the next big argument, but this is subjective. 
Moreover, changing datatype maps to datatype sets does not change the 
number of variables we need, unfortunately.

Please take a look at what I propose in my other email today, which 
retains almost all the text used in the CR version of Semantics (it 
keeps a lot of the simplified explanations) while using datatype maps at 
very few key places mostly in maths formulas.


>           -- Sandro
>> AZ.
>>> --------
>>> I believe this establishes sufficient connection between the 2004
>>> 'datatype map' terminology and the 2013 'recognized IRI' terminology
>>> to enable conditions stated in the 2004 language to apply clearly to
>>> the 2013 specifications. Do you agree?
>>> Pat

Antoine Zimmermann
ISCOD / LSTI - Institut Henri Fayol
École Nationale Supérieure des Mines de Saint-Étienne
158 cours Fauriel
42023 Saint-Étienne Cedex 2
Tél:+33(0)4 77 42 66 03
Fax:+33(0)4 77 42 66 66
Received on Wednesday, 11 December 2013 15:26:32 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:59:44 UTC