W3C home > Mailing lists > Public > public-sws-ig@w3.org > November 2003

RE: UDDI and semantics

From: Stephane Fellah <fellah@pcigeomatics.com>
Date: Wed, 19 Nov 2003 15:55:44 -0500
Message-ID: <8ED21571324EB145933ACCD22B86AC363C4012@bach.ncr.pcigeomatics.com>
To: "Farrukh Najmi" <Farrukh.Najmi@Sun.COM>
Cc: "Paul Denning" <pauld@mitre.org>, "www-sws-ig" <public-sws-ig@w3.org>

Najmum,

I dig out the OGC forum where I raise the issue for point 5. 

Here the extract of a post I sent:

"I have found a major weakness in the current RIM and ebRIM that could be really problematic in the future (especially for metadata service). The slot is not a subclass of registryObject. It is living outside the model. This is problematic because in the current state of the specs, you can not have twice the same slot name in a registryObject. It must be unique (difficult to ensure in a distributed environnement). If you are using metadata such as ISO 19115 and FGDC or others, you can be sure that the information have some overlapping naming. If you are using a namespace for each slot (basically consisting giving a URI for the slot), you could identify without ambiguity the term you are using. "


Best regards
 
Stephane Fellah
Senior Software Engineer
 
PCI Geomatics
490, Boulevard St Joseph
Hull, Quebec
Canada J8Y 3Y7
Tel: 1 819 770 0022 Ext. 223
Fax 1 819 770 0098
Visit our web site:  www.pcigeomatics.com
 


-----Original Message-----
From: Farrukh Najmi [mailto:Farrukh.Najmi@Sun.COM] 
Sent: Wednesday, November 19, 2003 3:35 PM
To: Stephane Fellah
Cc: Paul Denning; www-sws-ig
Subject: Re: UDDI and semantics



Stephane Fellah wrote:

>Paul,
>
>Farrukh Najmi sent this post today 
>http://lists.w3.org/Archives/Public/www-rdf-interest/2003Nov/0133.html. 
>Perhaps he has some additional links to existing work.
>
>I tried one year ago during the OGC Web Services 1.2 Initiative, to 
>come out with an partial ontology of ebXML registry model based on RDFS as a proof of concept. The model has been developed with Protégé (prior its support of OWL). The ontology is very rough because it is was developed at the time I was learning more about ontology, so forgive me if the model does not make sense in some part. I would certainly produce a more robust model in OWL today. You can find it at: http://gws2.pcigeomatics.com/ebRIM. The RDFS schema is at: http://gws2.pcigeomatics.com/ebRIM/ebRIM.rdfs. An example of instance is at: http://gws2.pcigeomatics.com/ebRIM/ebRIM.rdf (with dummy data).
>  
>
Stephane, this is very interesting work. I will study it in detail after 
a Dec 10 deadline I am focused on.

>I've learned the following lessons:
>
>1) I don't have to write any code or specific API to access the ebRIM 
>modeled in RDF !
>
>2) Using RDF query languages such as the one in Protégé or RDQL, maps 
>easily to the pre-canned query of ebRIM.
>
>3) It took me one hour to make the model and to get working queries on 
>the model. The model was working immediately (building instances and search the model) along the introduction of the concepts in the knowledge base.
>
>4) ebRIM overlaps a lot with OWL construct. Associations are modeled by 
>ObjectProperty, RegistryObject are Resource,ClassificationNode are OWL:Class.... InternationalString can be simplified by using the standard RDF construct xml:lang (used for example in rdf:label and rdf:comment).  My conclusion is that using OWL is more adequate and expressive than ebXML constructs. ebXML Business specific objects such as CollaborationProfile, ebXML Message concepts should be expressed in term of OWL concept and properties and avoid duplicating constructs existing in OWL.
>  
>
Very good summary. BTW the overlap is exactly what we would like to 
eliminate by aligning ebRIM with RDF and OWL in version 4 of ebXML Registry.

>5) Search are done on association name which do not have a unique id in 
>ebXML. This is not compatible with RDF approach. I don't know if this has been fixed since...
>  
>
Not sure which aspect you mean. All RegistryObjects including 
Associations have a unique id in ebRIM. That is how it has always been.

>I would be more than delighted that ebRIM is using OWL as a fundation 
>for the registry because it will simplify the complexity of the model and will ease the integration with other registry and semantic web technologies.
>  
>
That makes the two of us ;-).

While I have a lot to learn about RDF and OWL,  I know enough to say 
that they make the perfect foundation for ebRIM.
Conversely I feel that ebXML Registry has the potential for being to the 
Semantic Web what Web Servers are to the web today; ebXML Registry could 
be the "Semantic Web Server" for the semantic web.

The ebXML Registry TC could use more experts on RDF and OWL as active 
participants so they can help us get it right as we align with the 
semantic web in version 4.

If you, Paul or any one else on the public-sws-ig@w3.org list are 
interested and able to help then please let me know directly.
I look forward to collaborating with you and other colleagues in the 
semantic web services community. Thank you.

-- 
Regards,
Farrukh


|========================================================================|
|                                                                        |
| Going to XML 2003, Dec 8-12, Philadelphia PA?                          |
|                                                                        |
| Come see a very cool presentation and demo:                            |
|   "Epidemic Management Using OASIS ebXML, UBL and XACML Standards      |
|                                                                        |
| When: Wednesday, December 10 / 2:00 PM - 3:30 PM                       |
|                                                                        |
| Details at:                                                            |
|   http://www.xmlconference.org/xmlusa/2003/interopdemos_oasis.asp#4    |
|                                                                        
|| 
|=======================================================================
|=|
Received on Wednesday, 19 November 2003 15:55:48 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 16 March 2008 00:10:53 GMT