The WS-Agreement Standard as defined in the current proposal does not accommodate negotiation between service provider and requesters to yield an agreement. It is limited to simple offers and acceptances. However, negotiations between service providers and service consumers beyond the basic mechanism are required and are also part of the charter of the GRAAP working group. There are many mechanisms for negotiating agreements between parties for different purposes, having different sets of participants, different economic properties, different interactions patterns, and different information requirements to enter the interaction. Given the generality of the WS-Agreement scope, supporting a single negotiation mechanism would not satisfy the requirements of different application scenarios. Hence, the objective of this thesis is to devise a proposal to extend the WS-Agreement specification to facilitate the integration of arbitrary negotiation protocols into the WS-Agreement interaction. This includes facilities to enable a party to advertise the negotiation protocol it supports, the role that it can take in this protocol, expectations on other roles in the negotiation, and other meta-information that is necessary to start an interaction. This thesis analyzes the scope of negotiation protocols relevant for service level agreement establishment, derives a data structure for negotiation protocol meta-data, a protocol to exchange meta-data and initiate a negotiation protocol, and obtaining the result of the negotiation, the agreement. The new functionality extends the current WS-Agreement specification seamlessly. The thesis demonstrates the viability of the approach using scenarios of specific negotiation mechanisms.
Multi-Agent Systems (MAS) promise a new advance in distributed computing. In MAS autonomous software agents flexibly cooperate, coordinate, and compete to provide the desired function(s) of such a system. If some components of a MAS fail or do not provide the desired functionality, the system is expected to autonomously deal with these situations. It is desirable to reduce occurrences of such situations by selecting trustworthy cooperation partners before cooperating with them. This becomes even more desirable and important in an open MAS where arbitrary heterogeneous software agents, deployed by different parties, participate in the MAS, as it becomes more likely that these agents fail or try to exploit the MAS for their own purposes without reciprocation. In order to monitor agent behavior and enable selection of trustworthy cooperation partners, trust or reputation management services can be applied. As there is no central control in an open MAS and it is completely distributed these services themselves have to be distributed. This paper proposes a fully distributed reputation management service for open MAS based on peer-to-peer technology (especially distributed hash tables). A Java-based implementation of that service, which is intended as a plug-in for multi-agent platforms, is also described.