328 Chapter 9 Running a Tracy Agency
Current Status of Mobile Agent Toolkits
This is a result of the lack of any reference architecture for agencies for mobile
agents as well as the absence of an open and extendable implementation.
Therefore, each research group has to develop its own prototype. Because
of the complexity of developing a prototype and limited resources of the
research groups, this prototype is more a proof-of-concept implementation
focusing on a single research issue and leaving out the elementar y functional
components necessary for a full mobile agent toolkit. The research commu-
nity has to admit that such a reference architecture has been developed in
the area of distributed ar tiﬁcial intelligence in the form of the FIPA standard
[O’Brien and Nicol, 1998], for which a widely used implementation exists as
the Jade toolkit [Belliﬁmine et al., 2003].
We also face disparate perceptions of basic concepts of mobile agents, for
What should a mobile agent be from the programmer’s point of view: an
object of a speciﬁc type, which deﬁnes several basic functions for mobile
agents like communication and migration, or just any serializable object?
What level of communication is necessary: a simple one between agents
residing at the same agency or a complex one which also allows for
What level of security is necessary: one that protects hosts against only
malicious agents, or is it necessary to protect agents against malicious
What kind of mobility is necessary?
The disadvantageous consequence of these isolated islands of research is that
ﬁndings cannot be transferred between projects in the form of deﬁnite imple-
mentations (e.g.,software components that could be installed inothermobile
agent toolkits). Sometimes even the general research idea cannot be adapted
to another mobile agent toolkit because of the differences in basic concepts,
as previously described. Another deﬁcit is the number of different migra-
tion protocols that currently exist. Except for two toolkits (i.e., Aglets and
Grasshopper) that suppor t the MASIF migration protocol proposed as OMG
standard in 1998, it is virtually impossible to make two toolkits interoperable.
Although the MASIF standard provides a common migration protocol,
because of its complexity (which makes it difﬁcult to implement the complete