discussion:lazy, goals - A solution?
Johan Van Schalkwyk
jvs@iconz.co.nz
Tue, 29 Nov 1994 00:56:16 +1300 (NZDT)
Dear Tunespeople,
On Sat, 26 Nov 1994, Kyle Hayes wrote:
> I think we are making this _far too_ complex at this stage..
Abso-poso-lutely!
> One of the primary ideas of MOOSE TUNES (note that we still don't really
> have a name yet) is to be able to use a distributed network of processors.
Surely (after we've sorted out the basic basics) we should define _why_
we actually want a distributed network of processors. Surely not just for
the orgiastic pleasure of having a lot of processors doing it
simultaneously? I think that once we have defined a. general goals and b.
the goals of a particular transaction, then it will be a lot easier to
set limits on responses, define criteria for data integrity etc. We
should also be able to separate basic criteria (crc's on all
transactions?) from esoteric ones ("You sneaky bastard, you nearly
slipped one over me with that Trojan infested binary, but I got you this
time"?).
> It is all very well to talk about recovery, but you have to be able to
> tell just what you are recovering in the first place. Consider the total
> system state. [...] Let's approach these
> problems holistically. We have three subproblems: reliability,
> causality and efficiency.
What are our ultimate goals? (which will presumably will have a marked
effect on "what you are recovering", the "total system state", and your
three subproblems). I like to see computers as, initially, a tool, and
ultimately, a symbiosis. A system exists primarily to serve the needs of
the individual user - a tool. But, as we impart more "intelligence" to
the system, so it contributes more to the interaction, until it becomes
an "entity" in its own right, or with skilful software engineering, a
symbiosis that combines the beneficial features of human and machine
intelligence. Surely this is what we are aiming for, and distributed
processing, whatever, is merely a useful (and often complex and
frustrating) tool, rather than an end in itself?
Well, you asked for holism! (Trivia question, who invented the word)?
I would like to propose a concept. The concept is:-
*** The operating system as an organism ***
I do not think that this is entirely fanciful. I also believe that quite
apart from drawing trivial parallels between computer i/o and vision,
hearing, etc, we can still learn a lot from biological systems.
Principles such as, for example, those derived from analysis of the
iterated Prisoner's dilemma, have quite general application to both
biological systems and computer systems. Surely there are numerous other
principles that we can derive from biological systems?
If we see the user and "his" program as a symbiotic organism, and channel
all our energies into preserving the integrity of that organism, can we
not come up with something, as opposed to agonising about the finer
points of distributed systems?
Well?
JVS.