[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: copy protocol and object ravioli
From: |
Manor Askenazi |
Subject: |
Re: copy protocol and object ravioli |
Date: |
Fri, 14 Feb 1997 08:34:42 -0800 |
Copy: from "Programmer Parmesan" to "Environment Emmental"
----------------------------------------------------------
<< http://www.epicurious.com/db/dictionary/terms/e/emmental.html >>
Summary:
Objects in a simulation shouldn't have to support directly anything other
than what is required to model the subject matter at hand... There shouldn't
be any graphics code, memory management, internal logging facilities etc.
in the code for the actual model-agents.
Implication:
Swarm therefore needs a way for the programmer to declare which variables of a
given class are actually owned by every instance, and what message needs to be
sent in order to register a new instance with those variables which are not.
Finally, some standard mechanism for dealing with copy failure (due to lack
of resources or 'registration' failure) also needs to be defined.
Yeah, I think I can see why that would be a good thing...
I finally got it! ;-)
Manor.
PS> I also see why referring to the Java clone() solution was misguided:
the whole point in the clone() approach is that an object can copy
itself no matter where it lands (i.e. there is a minimal reliance on
the environment), whereas a SwarmObject always knows it is executing
in Swarm, so why not take advantage of that to alleviate some of the
hacking burden etc. etc.
PS> When will y'all have a draft API for this functionality?