FW: status of xsd:duration in OWL (and RIF and SPARQL) - ACTION-164: RDF WG

Hi Dan,

Here is one part of the discussion as a response to my request to consider date and time separately in addition to duration...

There is more... still digging

JP

-----Original Message-----
From: Evain, Jean-Pierre 
Sent: mardi, 8. mai 2012 11:18
To: 'Michael Schneider'
Cc: 'Ivan Herman'; Bijan Parsia; Ian Horrocks; public-owl-wg@w3.org; Public-Rif-Wg (E-mail); Peter F. Patel-Schneider; Sandro Hawke
Subject: RE: status of xsd:duration in OWL (and RIF and SPARQL) - ACTION-164: RDF WG

Dear Michael,

I appreciate your time and effort in trying to bring more background around the current situation.

I must say that I am growingly puzzled. This is definitely making me question my resolution to move for these technologies. If it cannot provide simple answers to simple questions, then maybe I am wasting my time. If I show your answer to some of my colleagues in my expert community, I may get some buying from those who have an academic background, but not from implementers (who are those who count to me in my daily business).

I believe that the semantics of time, date and duration are clear and I am surprised that they may be considered as being not mathematically univocally representable. For me there is nothing more semantically defined than a datatype bound to a particular format (and you'll always find cases where representation of date and time is ambiguous whether you use date, time or dateTime). A class instantiating such a datatype is also semantically defined in the context of a given ontology. Etc.

I believe working on the duration example would seem to answer part of the question but it is taking the easy way and in this particular case:
- you have taken an arbitrary time reference that is a second (what about tenth or thousandth of a second)
- you are facing the problem of defining the type of month according to its duration and resolving this as suggested looks interesting :--(
- then once you have calculated the value, how do you say on which basis it was calculated (e.g. how do you signal the unit unless it has to be seconds:--(, etc.)?

Why not simply reuse the xsd datatypes? That would solve all the above problems with a simple expression in a well defined format. What do I miss?

But you didn't really answer my question about expressing a start time in a video.  This is semantically perfectly clear and defined. I'd like to see an example of how this would be done and could be recognised as good practice by implementers.

I am not saying that you are wrong. You seem to have been thinking about it.. But I believe we do not live in the same world.:--) I am personally trying to be very practical and I realise that this technology may never fulfil my requirements to serve uniquely some theoretical purpose on improbable queries.

I am really wondering if this makes sense. Please convince me.

Jean-Pierre






-----Original Message-----
From: Michael Schneider [mailto:schneid@fzi.de] 
Sent: mardi, 8. mai 2012 10:39
To: Evain, Jean-Pierre
Cc: 'Ivan Herman'; Bijan Parsia; Ian Horrocks; public-owl-wg@w3.org; Public-Rif-Wg (E-mail); Peter F. Patel-Schneider; Sandro Hawke
Subject: Re: status of xsd:duration in OWL (and RIF and SPARQL) - ACTION-164: RDF WG

Hi Jean-Pierre!

Am 08.05.2012 09:25, schrieb Evain, Jean-Pierre:

> I understand the point which is being made being what is useful for reasoning or not.

Replace "useful for reasoning" by "required for the well-definedness of 
the semantics" of OWL 2 or RIF!

It is a basic technical requirement for the specifications of these 
languages that for every syntactically well-formed expression (aka an 
OWL 2 ontology or a RIF rule set), the semantic meaning can be 
determined by mathematical means. At a minimum, for OWL 2 and RIF, this 
means that it can always be determined whether an input ontology is 
satisfiable or not, or whether one given ontology entails another given 
one or not. Only in the cases of OWL 2 DL and its profiles, it is an 
additional requirement (by design) that there are reasoning procedures 
that are able to do these determinations in an automated way for all 
input, because these languages are required to be computationally 
decidable. But having a well-defined semantics is always needed. 
Clearly, if there are ontologies for which it cannot uniquely be deduced 
(mathematically) whether they are satisfiable or not, a reasoner cannot 
give the "right" reasoning result for them, because it cannot then be 
determined whether it's answer is right or not, or just one correct 
answer out of many.

To illustrate this problem, take the case of xsd:duration in its 
definition as of the time of finalizing OWL 2, where each literal of 
xsd:duration would essentially denote a pair (m, s) consisting of a 
certain number m of month plus a certain number s of seconds. Let there 
be two such durations:

     d1 := (2, 0)
     d2 := (1, 30*24*60*60)

Now, depending on what is meant by "a month", these two durations can 
represent either (i) the same value (if a month has 30 days), or (2) d1 
can be greater than d2 (if a month has 31 days), or (3) d1 is smaller 
than d2 (if a month has, say, 28 days = 4 weeks). I may well have missed 
a precise definition of "a month" in the (newest version of the) XSD 
spec, in which case the above example may be void. But if not, then it 
is clear that any OWL 2 (+xsd:duration) ontology for which the question 
of satisfiability depends on whether the above two durations are the 
same or not, or which of them is greater, does not have a uniquely 
defined semantic meaning.

An example for the need of being able to determine whether equality 
between two duration values holds or not would be an ontology with data 
enumerations consisting of duration values (denoted by "d1" and "d2", as 
defined above, but in a real ontology one would use their correct 
literal form, of course):

     :D a rdfs:Datatype ;
        owl:oneOf ( d1 ) .
     :dp a owl:DatatypeProperty ;
         rdfs:range :D .
     :s :dp d2 .

This set of axioms should be satisfiable if and only if d2 = d1, because 
only in this case, the object d2 of the property assertion (last 
statement) would denote an instance of the singleton datatype :D = {d1}. 
But if it cannot be determined whether d2 equals d1 or not, then it 
cannot be determined whether the axiom set is satisfiable or not.

An example for the need of comparison of two durations (greater or 
lesser than) could be constructed from the use of OWL 2 datatype 
restrictions.

> But does that mean that all other information is garbage?

Everyone can say everything about everything. But an OWL 2 or RIF or 
whatever language specification with a formal semantics at its core 
would, if not well-defined, IMO count as just that: garbage. :-)

Best,
Michael

-- 
..........................................................
Dipl.-Inform. Michael Schneider
Research Scientist, IPE / WIM

FZI Forschungszentrum Informatik
Haid-und-Neu-Str. 10-14
76131 Karlsruhe, Germany
Tel.: +49 721 9654-726
Fax: +49 721 9654-727

michael.schneider@fzi.de
www.fzi.de

..........................................................
Forschungszentrum Informatik (FZI) an der Universität Karlsruhe
Stiftung des bürgerlichen Rechts
Stiftung Az: 14-0563.1 Regierungspräsidium Karlsruhe
Vorstand: Dipl. Wi.-Ing. Michael Flor, Prof. Dr. Ralf Reussner,
Prof. Dr. Rudi Studer, Prof. Dr.-Ing. J. Marius Zöllner
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
..........................................................
------------------------------------------------------------------------------

**************************************************
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed.
If you have received this email in error, please notify the system manager. This footnote also confirms that this email message has been swept by the mailgateway
**************************************************

Received on Wednesday, 1 August 2012 17:48:27 UTC