Doing it better 1999-04-15 b
###

> I have (understandably?) made no programming progress with Tau, but have
> made some theoretical progress: I have a nice data structure which
> combines registers, stacks and memory blocks.

[snip]

> This has the nice property that any real VM (or even processor) can be
> described as a series of constraints on this idealised system (e.g. there

[snip]

A nice idea. But I don't think it has the practical application you
mention, because it is really much easier to write a Java class or
something. That's not to say it has no application at all...

> Obviously, something is missing: I need a similar description of
> "transformational" instructions that apply functions to bitstrings. I'd
> like a similarly nice model of them with the same property that the space
> complexity of the description is roughly proportional to the space
> complexity of typical code generated (or obeyed during interpretation) to
> implement it.

Again, I'd suggest that an interpreter is a pretty tight definition of a
language. Or maybe you want a looser definition? Or maybe just a
definition which is data and not code?

One option is to represent the data as a set of nodes, and hence the state
of the machine as a graph, and then the semantics as rewrite rules. I
think I may be able to generate a passable interpreter from such a
description. It requires work.

Alistair