KQML is a common format for the messages being sent in the communication process; each KQML message is like an object (in the sense of object-oriented programming), which has a performative (this can be thought of as the class of the message), and a number of parameters (instance variables).

The idea at the time was that the agents that used KQML to communicate could be implemented using different programming languages, and any information that agents had could be represented internally in different ways. The internal representation of each agent would generally be unique, but for communication purposes, agents treated other agents as if they had some internal representation of knowledge. This idea is called a virtual knowledge base (VKB).

Criticisms of KQML

Though development and distribution of KQML-based implementations did take place, the language did have a number of drawbacks:

• The meaning of performatives was very ambiguous, given in simple descriptions, which, if combined to the fact that the performative set was very large, limited the extent to which clear, efficient communication could occur. A whole set of performatives, called the commissives (or promissives), was missing, which was supposed to allow agents to commit to other agents. Hence, it was difficult to see how many multiagent scenarios could be implemented, negatively affecting coordination between agents.


• The basic KQML performative set was never tightly constrained, so different implementations of KQML were developed, that couldn't interoperate.


• The transport for KQML messages wasn't well-defined either, again making it hard for the communication process to occur as intended.


• The language was unsuitable for distributed problem solving/task sharing