Brian Rice
Sun, 10 Oct 1999 12:02:43 -0700

At 03:28 AM 10/10/99 -0300, Jecel Assumpcao Jr wrote:
>Rafael Kaufmann wrote:
>> I'd appreciate any thoughts, comments, suggestions, etc. that any of you
>> might have on the design and implementation and design of alisp (no
>> project website yet), a general-purpose, fully reflective language of
>> the Lisp family, and its accompanying interactive compiler. I intend to
>> start out by following the Scheme specification (R5RS), and
>> incrementally add features.
>That is certainly one way to do it. I prefer to start with a
>specification like that and then see what features I can remove...
>-- Jecel

I'd have to agree with Jecel here.  It seems against the grain of designing
reflective systems to give them more hoops to jump through to make
reflection workable.  This is certainly the case when you decide to
implement more features than are absolutely necessary, or include any
services that are not as simple to describe as possible for the core
language.  You'll just be adding unneeded complexity in the long run, IMHO.

To provide contrast, I am developing an interface to my Arrow prototype
from a Lisp interpreter written in Smalltalk.  Most of the design changes
involve simplifying the features of the parser, so that the more
complicated translation issues can be made by using the interpreter itself.
 But then, most of the issues involve conceptual shifts that programmers
haven't before addressed.  The point is that it seems to be a much simpler
feat to implement features within the reflective framework than to develop
the framework to support the feature set.