Re: Tool class (Long... *real* Long)

[Ken Cline]

On Thu, 20 Mar 1997, Sven N. Thommesen wrote:

> One question for Glen or anyone else who knows:
> if Agent B finds one of its methods invoked by
> *somebody*  and needs to know by whom, how does 
> it find out?
> 
> (I have in mind a situation where agent B finds it
> needs to invoke InvalidCombination or some such
> showstopping method, and wants to inform the world
> who made it do that, which I envision happening by
> agent B calling back to its caller asking it to 
> printSelf before saying "he did it!" and dying...)

Yeah, I'd really like to know this also!

I was thinking that the language should be extended by one
more (measly) pointer, namely, to go along with "self" and
"super", we'd have "caller".  (Maybe "caller" can be
extracted from the run-time stack?)

Are there languages that provide this?


> It occurs to me that the difference between a "tool" and
> an "agent" might become small indeed -- for example [to
> stay in Ken's bailiwick] if the "object" being "given" by
> Agent A to Agent B needs to fly from point a to point b
> before having its desired 'impact' ... In other words, if
> a user asks a tool to do something that will take sim
> time, the tool needs to make use of the scheduling
> apparatus and becomes, by Ken's definition, an agent !

Good point!  The line becomes even thinner... I guess
"tools" just aren't content being second-class agents. :)


> Perhaps we could call it CoolTools ? :-)

And I suppose only CoolAgents can use CoolTools ? :-)

> Instead of CoolTools, perhaps RoboAgents ?

I like that too!


Ken.

_________________________________________________________
Ken Cline                             address@hidden
SAIC                                 VOICE (410) 571-0413
Annapolis, MD                          FAX (301) 261-8427