Fwd: reuse of part-whole ontology in WSDL RDF mapping

From: Jacek Kopecky <jacek.kopecky@...>

Date: 21 January 2006 00:37:45 GMT

To: Alan Rector <Alan.Rector@...>

Cc: public-swbp-wg@..., rector@..., welty@...

Subject: reuse of part-whole ontology in WSDL RDF mapping

Alan, thanks for the info, I'm glad the part-whole ontology might move

on. I expect we will reuse it in some way in the WSDL RDF mapping, so

you can count on us for a review as well. 8-)

Can you please give me an estimate of the timing for this ontology? What

is the target status for the ontology? And especially, if we move the

WSDL RDF mapping to last call around the W3C TP in March, do you expect

the part-whole ontology to keep at most a step behind us, so that we can

reuse it without slowing down our already delayed deliverable? 8-)

And I have a guidance question as well:

We have currently a set of classes (e.g. Description, Interface,

Binding) and a set of properties (e.g. interface, binding) that are used

both for pointing from Description to its parts Interface and Binding,

but the "interface" property is also used to point from binding to the

corresponding interface (where there is no part-of relationship). So our

hierarchy is Description at the top which contains Interfaces and

Bindings (among others), and Bindings point to Interfaces (one each) as

well.

In order for us to reuse the part-whole ontology, the links between

Description and the lower level of Interface and Binding must indicate

the is_part_of_directly relationship. I can see 3 different ways in

which we can do this:

     1. when mapping from WSDL (XML) to RDF, we will generate both

        "interface" and "is_part_of_directly" between Description and

        Interface (and similar pairs of statements for the other part-of

        relationships)

     2. in places where a part-of relationship happens, we will replace

        the current named properties with is_part_of_directly, and we

        will keep our named properties (e.g. interface) to point from

        Binding to Interface

     3. we will split "interface" (and the same way for similar

        properties) into two properties - "contains_interface" and

        "interface", the first being subproperty of is_part_of_directly,

        the second used where "interface" is used currently but without

        the part-of meaning

I hope this makes sense. Personally, I don't like either of the options

- the first one shows no relationship between "interface" and

"is_part_of_directly"; the second makes any query for all Interfaces go

one level deeper (to the type of the object of is_part_of_directly), and

the third introduces the two properties, both of which should really be

a single "interface".

Do you have any comments on which approach might be preferable or if 

I indeed missed something better?

Best regards,

Jacek

On Fri, 2006-01-20 at 10:04 +0000, Alan Rector wrote:

Jacek

As far as I know the note is ready to go modulo a few minor tweaks.   

I think Chris Welty 'has the token'.  There was a delay when his  

machine failed at the Face-to-Face in November, and I suspect  

holidays and other things have meant the document hasn't emerged.

If there is a problem for Chris, I can probably make the few  

necessary changes next month, but he had some alternative examples in  

mind.

No changes in substance are planned.

Regards

Alan

On 18 Jan 2006, at 19:20, Jacek Kopecky wrote:

Dear SWBP WG, 8-)

within WS-Description WG we are working on an RDF mapping for the WSDL

components, and we are considering using your Part-whole ontology [1].

Can you please let me know about the current status of this particular

ontology and your plans for it?

We will possibly want to go to Last Call with the RDF mapping  

around the

Tech Plenary in March, so if the part-whole ontology is planned to be

delivered significantly later (if, indeed, ever), we would not be able

to reuse it.

Best regards,

Jacek Kopecky

[1] http://www.cs.man.ac.uk/~rector/swbp/simple-part-whole/simple- 

part-whole-relations-v0-2.html

-----------------------

Alan Rector

Professor of Medical Informatics

Department of Computer Science

University of Manchester

Manchester M13 9PL, UK

TEL +44 (0) 161 275 6188/6149

FAX +44 (0) 161 275 6204

www.cs.man.ac.uk/mig

www.clinical-esciences.org

www.co-ode.org