Infrastructure for Secure Interoperability of Agents

摘要

Introduction Building distributed applications is difficult. Therefore, it is hardly surprising that in spite of all the hoopla surrounding the Internet and distributed computing, truly distributed applications are few and far between. The problem seems to be with the tools available to developers of distributed applications. For example, the most widely used mechanism for distributed computation is the remote procedure call (RPC), the first implementation of which dates back to the early 1980s. Typically, a remote procedure call is executed on a server on behalf of a client (the so-called client-server model). It is hardly surprising therefore that most distributed applications today are exclusively based on the client server architecture. A lot can be (and has been) accomplished with this architecture, as exemplified by the World Wide Web and HTTP, a protocol that implements RPC. However, the client-server model has a number of limitations. There are problems of fault tolerance, load balancing, survivability, dynamic reconfiguration, rollover recovery, and distribution of control. Attempts in the past to break through this bottleneck have had only limited success.