| Communication is a central problem in
distributed systems. Without the ability to communicate,
a set of agents would be a merely a collection of
isolated systems. All classic distributed systems
problems, such as coordination and distributed reasoning,
depend on an underlying communication framework. Before
communication can take place, there must be a known
destination. KNS, like the KQML ANS scheme, adds a
communication layer in which symbolic names are mapped to
actual internet addresses. In addition, however, it
offers more advanced support for dynamic group formation
and disbanding, and a more abstract notion of agent
identity, independent of any particular name or globally
unique identifier. |