class, prototype, meta
John Albert Fehr
jfehr at sprintmail.com
Sun Mar 1 22:02:37 UTC 1998
wirth at almaden.ibm.com wrote:
>
> While we're on non-CS points of view, let me ask a new question about
> method bindings. I'm really interested in biological analogs (and was much
> taken by the ones you mentioned in your OOPSLA keynote). I've been trying
> to figure out how one might do method dispatching for prototypes that are
> floating around a distributed, sometimes-connected environment (as span the
> set of electronic gadgets I have in my pocket, briefcase, etc.). Cells
> have receptors on their cell walls that signal what messages (proteins)
> they can receive. Can we use a similar mechanism for objects? (The answer
> to which leads to more questions about where the methods are stored, how
> they're inherited, etc. The fantastic redundancy of having the full
> "source code" in the DNA in each cell is a marvelous thing.) Comments?
> Pointers to other writings on computational vs. biological systems?
>
> Mike
i don't have any "pointers to other writings" but i want to comment.
i liked what Alan had to say about the region between a prototype and
a "kind" -- that something starts out as a sketch and evolves into a
"kind of" something. i've wondered if a behavior for a prototypes or
classes couldn't die if unused, like entrophy. one could prototype a
method definition, allow it to keep track of its invocations and
versions, and the most frequently sent definitions could persist
across the temporal prototype-to-kind evolution phase. those methods
definitions that are unsent would either die or remain in the prototype
for further refinement -- excluded from the emerging "kind."
furthermore, in an local image, unwanted (unsent) behavior could be
treated like unreferenced objects -- eligible for garbage collection.
if the behavior is ever needed, its definition might be fetched from
another local image, compiled for that platform and evaluated -- as
in a set of local images "floating around a distributed, sometimes-
connected environment" or a kind of cyber-image. one could measure
when a prototype is a more-stable "kind" by the frequencies of its
instantiation and the utilization of its respective behaviors.
excuse me if this has all been discussed. i just subscribed to this
forum. i liked your question about method bindings and wondered if
dispatching might be assisted by somehow tracking behavior
invocation, -John F.
***********************************************************************
John Albert Fehr
Ranch to Market, Inc.
PO Box 162404
Austin, Texas 78716-2404
(512) 472-6726
jfehr at sprintmail.com
***********************************************************************
More information about the Squeak-dev
mailing list
|