Ken Evitt
Wed, 28 Apr 1999 22:28:38 -0400

>I propose the common goal is this, first to define a framework that
>defines concepts.  Basically something that has at the
>very least the exact same GOALS as Brian's arrow system (myabe his is the
>solution, maybe not, we don't know yet but the more I read the nicer it

So the primary goal of the Tunes Project is now to define a complete
epistemological theory? I don't think you could define a framework to handle
concepts that could be implemented on a computer--because the set of
concepts is not closed, there are concepts that are not defined in terms of
other concepts--many that cannot be reduced to more fundamental
concepts--because concepts are fundamentally based on percepts. Somewhere
among concepts there are concepts that can only be defined ostensively--that
can only be percieved. I think your plan suffers from the fact that you try
and seperate theory and practice--you want to formalize a system to
represent ideas--but how exactly would you do it and how the hell would it
work? And why do you need this framework and not merely a fully reflective
architecture? Do you need to write an AI, universal meta-translator and a
magic box to make computing fundamentally better?

>The plan to me looks like this;
>Once we have that, we have a way of the computer understanding something
>like an API or any program, and with that understanding it can do things
>like verify the program function, glue API's together automagically etc.
>Once we have the specification down in stone, we can try to impliement it.
>Once that's impliented using currently available tools on an existing OS
>(probably linuux since most of us have PCs and linux has lot's of good
>tools regarding language etc.).  Once that is working we might want to
>design meaty OS chunks with our system, things like filesystems,
>schedulers, a sort of specification of the hardware we are working on (so
>our system can handle OS specific stuff like memory paging etc. for
>us); under our system things should be dynamic so that one can change
>features of any of the components we designed in real time, as no one
>general component will be the most effective for all jobs.  Then we can do
>games, word processers etc.

It sounds to me like your goal is to design a logic system where you give it
a list of declaritive statements about whatever it is that you're talking
about and the system produces correpsonding imperative statements--you want
to specificy what should happen and let the system handle the details, yes?
If so, I don't see how in hell you'll ever accomplish that.

When you say that the computer has a way of understanding something like an
API or a program, do you mean in the exact same sense as a human being
understands an API or a program? The computer could positively identify the
exact function of any program by itself? I don't see how it could do this.
You mention designing meaty OS chunks, but I thought this was futile--you
said the computer should dynamically determine things like scheduling
algorithms to be optimal--how are you going to do that? Your computer sounds
like a magic box--all you have to do is tell the computer the problem by
specifying it and out pops the answer. How will this be accomplished?

>And that is what I believe is a plan.
>Objections..what are they?
>Tunes won't evolve into an's really a mix up of words.  What we
>want is a framework on how to deal with
>concepts, then to sort of bootstrap and immerse the computer in that
>framework will consitiute an environment or an it what you will.
>By designing our OS with our meta framework, makes an OS that is
>beutifully created, dynamic and sorts of other nice words...everything
>(memory manager, scheduler, vmem etc.) is a component of the system,
>and the OS becomes a set of concepts that are connected to most other
>programs.and it can "understand" the programs that are in it(be able to
>prove functionality), just how
>everything is represented will ultimately be a choice of the user, as no
>one context fits all.
>So this is my solution, can we agree on it?

What words is the mix up over?

Isn't an operating system a framework that deals with computing concepts?

I don't see why someone can't be working on a boot-strapped system. We could
define the HLL by extending Scheme or designing it from scratch, or
whatever, and then we could implement that in the working system and because
the language is reflective and the basic system is designed to be a part of
the language/OS, gradually the language can be used to redefine things in
cleaner ways and then the process of porting the system becomes easier
because the level of abstraction is no longer at the hardware level.

I think I agree with you on the idea that Tunes is not an operating system
on top of which other things are defined but is the system or context (or
environment) that gives meaning to the computing objects which the user

I tried to read Brian's paper but it was so densely populated with what
seemed "meta-babble" that I stopped. Though everything else Brian has
written has been clear enough.

-Ken Evitt