RFC: HLL Architecture outline
Brian T Rice
Sun Mar 16 04:40:02 2003
On 15 Mar 2003, Joerg F. Wittenberger wrote:
> Brian T Rice <firstname.lastname@example.org> writes:
> > Anyway, here's the bird's-eye perspective:
> > We should organize the Tunes project around language architectures, not
> > particular languages themselves. In this sense, "HLL" and "LLL" could come
> > to mean "High-level Language Library" and "Low-level Language Library",
> what's the difference actually? If I abstract from "syntax" when
> talking about a language, then there is not much of a difference
> anymore. As such good point (as above).
Because the semantics don't have to be a single-agreed-upon-thing either,
in addition to having other syntaxes. I could define different
implementations of the same data structures which had slightly different
semantics: one for a functional style of context, and the other for a
context where mutation is expressed directly.
So then the only real difference is that one area of the system deals only
with high-level concepts and the other deals with the platform-specific
> > consists of: input-runtime-output corresponding to Read-Eval-Print from
> > Lisp, and at the basic level consists of Parse-Expand-Search-Dispatch-
> > Rewrite-Response.
> Shouldn't that Parse-Expand-Search-Dispatch-Rewrite-Response level of
> detail already go into "low level desging". I observed strong
> similarities between the above page and the low level design of
> Askemos (minus a some details).
But it's not /low-level/. Low-level has to do with code-generation and
bit-munging strictly. This simply deals with components at a
meta-linguistic level, and TUNES is that much more useful because we can
express it this way.
The reason that this works is that the HLL and LLL are expressible as
languages as well, and that this meta-level is then just inter-language
translation of a simple kind. And, yes, I have considered moving out that
page to the meta-translator area.
Brian T. Rice
LOGOS Research and Development