Re: message to RDF entailment implementors

On Nov 1, 2013, at 4:48 AM, Eric Prud'hommeaux <eric@w3.org> wrote:

> * Pat Hayes <phayes@ihmc.us> [2013-10-31 10:50-0500]
>> Reads nicely. I wonder if the immediate-inconsistency of illtyped literals should be spelled out in a bit more detail, and the notion of recognized datatype be expanded slightly. Minor wording changes along those lines suggested here.
> 
> Is there any advice to be offered to folks who don't understand the implications of an inconsistency?

An inconsistency entails anything, even another inconsistency. 

> I can imagine different classes of readers wondering if this change applies to them:
> 
>  SPARQL users who are unaware that they are just using graph entailment: SPARQL says that querying for {<s> <p> ?o FILTER (datatype(?o) = xsd:integer} over { <s> <p> "ab"^^xsd:integer } will get you {(?o->"ab"^^xsd:integer)}

That seems OK, since this inconsistent graph would entail anything. But in any case I don't think that SPARQL has ever set out to check consistency in the graphs it queries, right? 
> 
>  Same users who are using an engine doing RDF/S or DL entailment asking questions which don't require that triple with the malformed object.

Then it shouldnt have any effect on them, I would guess. But even with the 2004 semantics, a graph can be RDFS-D inconsistent. Did SPARQL ever handle this case "logically", ie by saying that all queries are satisfied by an inconsistent graph? For example, using the 2004 semantics

:a :p "ab:^^xsd:integer .
:p rdfs:range xsd:integer .

is unsatisfiable, so it RDFS-{xsd:integer} entails, for example,

:b :q :c .

But if you had queried ?x :q :c against that first graph, you would NOT have got a match, right? 

> 
>  Same again, but who are touching the wretched triple (sameAs? cardinality? i dunno).

Makes no real difference.

Pat


> 
>  Users of OWL API, etc.
> 
> 
>> On Oct 30, 2013, at 3:11 PM, Peter Patel-Schneider <pfpschneider@gmail.com> wrote:
>> 
>>> One thing that I think would be nice to send out to implementers of RDF entailment systems is something saying what has changed.   Here is a draft of the information. Comments are welcome.
>>> 
>>> peter
>>> 
>>> 
>>>  Entailment-visible changes for RDF 1.1 (informative)
>>> 
>>> Most of the changes between RDF and RDF 1.1 do not have any effect on
>>> implementations of entailment, but there are a few minor changes.
>>> 
>>> The sequence in which the versions of entailment are defined has changed.
>>> Datatype entailment is now defined on top of simple entailment, and then
>>> RDF and RDFS entailment are defined on top of datatype entailment. Datatype entailment refers to a set of 'recognized' datatypes.  RDF
>>> entailment has two required datatypes xsd:string and rdf:langString which must be recognized, but
>>> this doesn't appreciably add to RDF entailment as these two datatypes
>>> replace plain literals. 
>>> 
>>> Literals formerly described as plain are now divided into xsd:string literals for plain literals
>>> without language tags and rdf:langString literals for plain literals with
>>> language tags. Thus, all literals have a type and there is no need for an implementation to have
>>> separate data structures for plain literals and datatyped literals,
>>> although rdf:langString is a special datatype as it has a language tag in
>>> addition to a lexical form and thus requires special treatment.  The zero
>>> Unicode character is not allowed in xsd:string, but was allowed in plain
>>> literals, so there is a minor change here.  Implementations that have a
>>> special internal data structure for plain literals might not need to
>>> otherwise change.
>>> 
>>> One change that does affect entailment is that graphs containing invalid literals (e.g.,
>>> "a"^^xsd:integer) are immediately inconsistent for recognized datatypes, even in sub-RDFS entailment regimes. 
>>> 
>>> There is a list of XML Schema datatypes that are deemed suitable for use
>>> within RDF.  They are all optional except for xsd:string.
>>> 
>>> The rdf:XMLLiteral datatype is now optional.  rdf:HTML is a new optional
>>> datatype; implementation experience and illustrative tests are requested.(Note also that this has at-risk aspects concerning DOM4 normalization.) 
>>> rdf:PlainLiteral is a newish optional datatype; implementation experience
>>> and illustrative tests are requested.
>>> 
>> 
>> ------------------------------------------------------------
>> 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
>> 
>> 
>> 
>> 
>> 
>> 
>> 
> 
> -- 
> -ericP
> 
> office: +1.617.599.3509
> mobile: +33.6.80.80.35.59
> 
> (eric@w3.org)
> Feel free to forward this message to any list for any purpose other than
> email address distribution.
> 
> There are subtle nuances encoded in font variation and clever layout
> which can only be seen by printing this message on high-clay paper.

------------------------------------------------------------
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 Sunday, 3 November 2013 05:57:29 UTC