Agents and Jini
Posted by Bill Venners on January 12, 20100 at 10:51 PM
> I very much liked the "A Walk Thru Cyberspace" article.
> I thought that modelling cyberspace on virtual places was a good
> idea and would allow everyone to use cyberspace without prior
> training and would be intuitive. "Places" are important building
> blocks in mobile agent systems as well. These places could host
> electronic commerce activities on behalf of cyber businesses.
> I really liked the article.
On the subject of agents, I wrote two articles back in early 1987
about IBM's aglets. One article described the architecture of
aglets. The other article attempted to explain why we should care,
to get at the point of agents. I really had some difficulty
seeing how sending the state along with the code (which is
what makes something an agent) gave much advantage over sending
just the code (such as a Java applet). I listed some examples
that had been kicked around by agent enthusiasts, but was
luke-warm in my own interpretation.
As I worked with Jini, it occurred to me that Jini service
objects are agents, because the state is being sent along
with the code. They also go to more than one place. Normally,
the service provider sends the service (and other) objects up
to the lookup service, which later sends the same objects out
It finally dawned on my then that a major point of agents could
simply be to raise the level of abstraction for network
programming, as Jini does, from the bits and bytes of
protocols, to the services offered by objects.