Abstract

In recent few years, the use of "policies" in the management of the behavior of large-scale, dynamic or distributed systems is gaining importance. In the policy research area, how to capture policies is the first problem scientists need to solve. Policies can be specified in three ways: a Policy Definition Language (PDL), a formal logic-based policy representation language or an informal natural language. Generally, an editor and a policy management tool are used to input these specifications. Because both the PDLs and the formal logic-based policy representation languages are very complex and more suitable for machine communication instead of human communication, their usages are limited to a small group of professional people. In this thesis, two levels of policy specifications are introduced. A Restricted Natural Language (RNL), as the high-level policy specification, is used for human-computer interactions. Compared to the PDLs and the formal logic-based policy representation languages, the RNL is both easy to use and easy to learn. It enables a wider group of end users, especially the novice users, to participate in stating the policies. A logic-based formal representation, as the low-level policy specification, is then used for further machine-oriented processing in policy-based systems. A methodology of how to construct such a RNL is also presented and the methodology is applied to two different cases. In the implementation part of the thesis, a RNL user interface (RNLI) named the Virtual Thesis Office (VTO) System is also developed and presented