Evaluation and Implementation

To test our approach to DAML-S we implemented two systems with very different characteristics that take advantage of DAML-S and the Web Service architecture described above. The first one is a B2B application in which a Web Service looks for business partners and automatically negotiate business agreements with them. The second application is a B2C application in the travel domain in which a Web Service that functions as personal assistant of a user organizes a trip to a conference by booking a trip to the conference verifying availabilities with the user schedule stored on MS Outlook.

The first system describes a B2B application in which a Web Service that is given the task of assembling computers looks for providers of computer parts. The architecture of the Web Service is described in figure . The Interface Agent provides an operator with a way to interact with the planning agent and to compare options of combinations of business partners schedules costs and so on. The Planning Agent employs the planning scheme described above to achieve the goals proposed by the operator, in our case to find providers of computer parts and organize a supply chain that meets cost and time limitations. To achieve its goals the planning agent queries the Matchmaker for potential parts suppliers, then it uses the Toshiba and Fujitsu financial services to verify the likelihood that the suppliers will not bankrupt during production time affecting the whole business. Finally, the planning agent contacts the suppliers to negotiate schedule and costs.

Figure: Description of the system for scheduling a trip to the DAML PI meeting

The challenge of this system is to support interaction between Web Services provided by very different organizations geographically spread which we could orchestrate only on the bases of DAML-S information. Furthermore, it allowed us to experiment with suppliers selection using information that is erogenous to DAML-S. It is unfeasible to expect that DAML-S Profiles will contain all the information that the requesters will ever need about the provider, in this case the requester, i.e. the Planning Agent, uses two financial services to gather information about its providers before contacting them.

The Matchmaker used in the system is a DAML-S enhanced UDDI, it uses a freely available UDDI server to store DAML-S advertisements using the encoding described in [11] allowing for capability matching in UDDI.

The organization of the second example is displayed in figure ; the goal is to book a trip to a conference, namely the DAML PI meeting. We assume that the organizers of the meeting publish a Web Service which provides information about the meeting, such as time, location, talks, participants and so on. Through the RETSINA Calendar Agent [14], the user plans a trip to the conference. The Calendar Agent verifies availability checking on the schedule of the user stored in MS Outlook, and then uses the same Matchmaker used in the previous system find airlines, car rental companies and hotels. Finally, uploads the schedule of the trip in Outlook.

Figure: Description of the system for scheduling a trip to the DAML PI meeting

This example extends the previous one by using complex process models that we used to implement the Web Service as well as to be loaded dynamically to control the interaction on the client side.