RE: Agents components as services

Mr. Abela (et al),

   We at Daniel H. Wagner Associates are working on an ontology-driven
process for creating software agents from reusable components.  Under
government contracts with several agencies, we have spent the last couple of
years honing our approach, and I can assure you that it is feasible.  We
have developed a prototype tool called the AgentWizard, which provides
several types of interfaces for end-user combination of components into
working agents.  These interfaces include the Agent Builder, which provides
standard drag-and-drop of Java Beans from a component toolbox onto a palette
for visual component combination, as well as more sophisticated Requirements
Wizards, which are designed toward specific knowledge domains and step the
end-user through a series of questions designed to result in a full
definition of agent behavior for matching against rule-based operations.  As
you suggest in your email, the integration possibilities are staggering --
hence the support we've received from government agencies.
    If you would like to discuss this approach further, please contact me.

Regards,
James Eanes
Senior Analyst
Daniel H. Wagner Associates
2 Eaton St. Suite 500
Hampton, VA 23669
Phone: (757) 727-7700  Fax: (757) 722-0249

-----Original Message-----
From: www-ws-request@w3.org [mailto:www-ws-request@w3.org]On Behalf Of
Charlie Abela
Sent: Saturday, April 13, 2002 11:10 AM
To: www-ws@w3.org
Subject: Re-:Agents components as services


  I am working on the following idea and would like some feedback from the
group.



  The idea is to create DAML ontologies to handle the composition of
software components. I am thinking of having an ontology that describes
rules for the integration process and one that describes the composition of
java processes. In this latter I am considering describing java classes and
methods. In an instance of this ontology then, would be defined a particular
process, which when interpreted would result in something of this sort:

              ClassA  x = new ClassA();

              output = x.Method1(input);

              ClassB y = new ClassB(output);

  Etc…etc….



  This in my opinion will give the programmer flexibility in integrating new
software to an application, such as a software agent. One possible use would
be for example when an agent needs to communicate with other agents but it
does not have this ability because it can communicate with only one ACL. By
integrating new software this ability could be obtained.



  Does this idea make sense…. is it feasible? Or maybe there is some flaw in
the argument I am bringing?



  Any feedback or idea of previous/current work in this area is greatly
appreciated.



  Regards,


  Charlie