Re: [soa-rm] RE: Resolving Various Policy Languages with Ontologies

[Frank McCabe <frank.mccabe@us.fujitsu.com>]

I second this. Although it has been a while since I had to re- 
engineer a toilet!

As you might have guessed, I have a slightly eclectic view of  
ontology. I feel fairly firmly that the currently fashionable  
approach of defining ontologies by means of classifications is  
essentially broken; for 2 main reasons:

1. You cannot separate a concept network from a purpose. For example,  
the set of concepts that are important in building furniture is not  
the same as the set of concepts that are important for designing  
office layouts -- even though both may mention chairs and tables in  
some detail.

  Thus, mapping between ontologies becomes a task bridging activity.  
This is something that the SW community does not yet seem to have  
understood. <duck/>

2. The presupposition of many ontology notations is that the world is  
composed of objects and it is the ontology engineer's task to  
correctly classify those objects into conceptual hierarchies.

  However, even objectification (the process of identifying objects  
from a continuous analog data stream) is already a non-trivial  
activity involving a lot of interpretation and that cannot be  
separated from the purpose of the identification. There is a lot of  
neurological research that supports this perspective.

For our work, I think we ought to be careful and judicious in  
invoking concepts such as Ontology. We should focus on the  
requirements of describing service qua service; and leave the rest to  
another day.

Frank


On Oct 11, 2005, at 9:23 AM, Ken Laskey wrote:

> I remember making some notes on the ontology section that I wanted  
> to pass Frank's way but I've haven't gotten to that yet.  (However,  
> I did just fix a nonfunctioning toilet, so I have my engineering  
> accomplishment fix for the day :-) )
>
> Anyway, I believe ontologies can be a very powerful tool in  
> bringing together concepts from independent but overlapping domains  
> (subdomains?) but we need to be careful to specify concepts  
> (including needs) in the RM portion of or work without specifying  
> solutions.  Thus, our list of what would be discussed in an RA  
> portion continues to grow.
>
> Ken
>
>
> On Oct 11, 2005, at 11:50 AM, Duane Nickull wrote:
>
>> Post from Danny Thornton:
>>
>> (he mentions the "O" and "S" words)
>>
>> ;-)
>>
>> -----Original Message-----
>> From: Danny Thornton [mailto:danny_thornton2@yahoo.com]
>> Sent: Monday, October 10, 2005 10:26 PM
>> To: Duane Nickull
>> Subject: Resolving Various Policy Languages with Ontologies
>>
>> Hi Duane,
>>
>> The following is an e-mail dicussion I would like to
>> have with soa-rm group:
>>
>> I have been reading WD-SOA-RM-09 to get an idea of the
>> terminology/concepts for resolving various policy
>> languages in a service oriented architecture. Section
>> 2.2.3.2 of WD-SOA-RM-09 discusses the limits of
>> description.  Section 2.3.1.2 states that an ontology
>> can be defined to interpret strings and other tokens
>> in the data.
>>
>> In the discussions I've had about resolving various
>> policy languages in an SOA, I've hijacked the ontology
>> concept and applied it as a general concept for
>> resolving differences in policy languages.
>>
>> For example, if I have a service that uses XACML
>> policy and another service that uses EPAL policy, I
>> could resolve the differences between the two policy
>> languages using an ontology for both policy languages
>> at the policy decision point.
>>
>> For section 2.3.1.2 of the WD-SOA-RM-09, does anyone
>> have any thoughts on expanding the concept of
>> ontologies beyond the service description's data
>> model?
>>
>> Danny
>>
>>
>>
>>
>>
>> __________________________________
>> Yahoo! Mail - PC Magazine Editors' Choice 2005
>> http://mail.yahoo.com
>>
>
> ---
> Ken Laskey
> MITRE Corporation, M/S H305     phone:  703-983-7934
> 7515 Colshire Drive                        fax:        703-983-1379
> McLean VA 22102-7508
>
>
>