The biological cell (was: Erlang)
Alan Kay
Alan.Kay at squeakland.org
Sat Nov 15 15:01:05 UTC 2003
Thanks Marcus --
These are both good observations.
This is why I thought having objects be actively looking for messages
was a good idea, and I really agree with why Linda is a nice model
(perhaps having the matchers be associated with active objects is
even nicer). In the early implementations of ST there just didn't
seem to be enough computing power (and brain power in our heads) to
make some of the 60s ideas (e.g. Dave Fisher's notions about control
structures as transmuted to objects) work in a practical manner.
Gelernter's later Linda ideas are quite suggestive. David Reed's
ideas in his thesis and Croquet are also very suggestive of what more
modern object models might be like. BTW, they include the notion of
continuous rather than discrete time (as did the first Simula), and
that processes are really more like mathematical entities that are
approximated "behind the curtain".
The "strength of message" idea has been tried a few times. One of the
earliest observations about Sketchpad was that it didn't have "an
inverse square law" and that such a law would really be useful.
Borning experimented with confinement (which worked, but perhaps was
too total), and the SONY Research Corporation OSs of Mario Tokoro
(called MUSE early on, and now Aperios) went much further to a
field-theory of message propagation. I certainly have liked this
model.
Cheers,
Alan
At 2:29 PM +0100 11/15/03, Marcus Denker wrote:
>Am 15.11.2003 um 10:21 schrieb Diego Gomez Deck:
>>
>>One problem I see in the current receiver/message model in Smalltalk is:
>>The receiver is so hard-coded.
>>
>>We always have an explicit receiver. A type of "message in a bottle"
>>(without an explicit receiver) is something to experiment. A message
>>without an explicit receiver can different answers from different
>>receivers (did you try to shout in a place full os people ;))
>>
>Yes. If you look at biology, then most of the "messages" (called signals)
>are never sent to a specific reveiver, but just "dumped in the environment",
>were all possible reveivers can pick them up or ignore them.
>
>So these signals are much more like the "pheromone trails" in StarLogo
>or a Linda Tupple-space than squeak message sends...
>
>Another thing we have no analogy for is the "strength" of the message: In
>biological systems, many processes are guided not by "got a signal/got
>no signal", but by the strength or duration of the signal. e.g. gradients of
>signals play a *huge* role in embryo development.
>
> Marcus
>
>--
>Marcus Denker marcus at ira.uka.de
--
More information about the Squeak-dev
mailing list
|