W3C home > Mailing lists > Public > public-owl-comments@w3.org > March 2009

Re: [LC Response] To Marijke Keet Re: haskey clarification

From: Marijke Keet <keet@inf.unibz.it>
Date: Thu, 05 Mar 2009 10:45:53 +0100
Message-ID: <49AF9F51.8000707@inf.unibz.it>
To: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>, public-owl-comments@w3.org
Dear Peter,

Thank you for the response. However, I disagree with the argumentation.

"OWL systems" (any software system implementing DL languages, as far as 
I'm aware of) use so-called internal identifiers, i.e., system generated 
identification, therefore real keys (the ones with at least 1:1, and 
also mandatory if unidentified objects are not allowed to linger around 
in the system)--also called semantic keys or natural keys--do not have 
to be specified by the modeller and then modelling object uniqueness 
with 1:n (just inverse functional) suffices. UML has internal 
identifiers as well. ER, EER--where the notion and terminology of keys 
comes from--and ORM have user-defined natural keys (with mandatory 1:1) 
instead. Using the same terminology for different things easily leads to 
confusion and incorrect assumptions and this could have a knock-on 
effect in OWL 2 applications the userbase are, or is going to, develop; 
if the WG wants to risk that, that's your judgement call.

As for the passport example: if one assumes individual countries with a 
passport system for their citizens, and they would record their emitted 
passports only, then it suffices for identification because there is a 
1:1. If, on the other, hand one develops a system that should record all 
passports a person has or integrates the databases about passports 
across the world, then passport number is not adequate for 
identification of a person (the same holds for another often-used toy 
example: social security numbers). Such discussions are in the realm of 
conceptual modelling where finding a sufficient condition within the 
limited application area is the aim, and, arguably, ought not to be  
part of ontology design where one tries to represent information of a 
subject domain instead [1], but that is a separate topic.

[1] what the necessary and suffient conditions are to identify a person, 
I do not know either, but I think you will agree with me that even if we 
both do not have passports with numbers, we are still persons and we can 
distinguish you from me in some way by using properties x, y, z that 
could then be represented in the ontology.

Best regards,

Peter F. Patel-Schneider ha scritto:
> Dear Marijke,
> Thank you for your message
> <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0062.html>
> on the OWL 2 Web Ontology Language last call drafts.  
> The OWL working group realizes that keys are often functional (and thus
> 1:1).  OWL in general does not have a requirement that properties are
> functional.  Non-functional keys are useful in some cases, for example,
> passport numbers are keys for people but quite a few people have more
> than one passport number (because of multiple citizenship).
> In any case, as you say, properties can be made functional in OWL,
> either locally or globally, and this provides for functional keys.
> Therefore, the OWL working group does not intend to change OWL 2 or to
> make any changes in the OWL 2 document in response to your comment.
> Please acknowledge receipt of this email to
> <mailto:public-owl-comments@w3.org> (replying to this email should
> suffice). In your acknowledgment please let us know whether or not you
> are satisfied with the working group's response to your comment. 
> Regards,
> Peter F. Patel-Schneider
> on behalf of the W3C OWL Working Group 
Received on Thursday, 5 March 2009 09:46:35 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:40:21 UTC