Syntactic extensions to Scheme

[David Bartley <Bartley%CSL60%ti-csl.csnet@CSNET-RELAY.ARPA>]

One of the issues shunted aside a year ago was the question of how to
provide users with the ability to make syntactic extensions.  I think
it would be fruitful to have some calm discussion on this topic in the
RRRS-AUTHORS mailing list.  I doubt if we're ready to standardize on
any particular approach, but it would be helpful if we all knew more
about the alternatives and people's experiences with them.

I'm particularly interested in experiments such as T's with extensive
support for syntax tables.  What are their strengths and weaknesses?
What do they cost in terms of compiler or runtime complexity?  Do
users make effective use of them?  How flexible are they---would they
suffice for major language changes?

Should syntactic extensions be specified entirely in source terms, or
should the user have access to internal representations?  Should
simple source-to-source optimizations be communicated to the compiler
using the same mechanism, or is something else more appropriate?

Regards,
David Bartley
-------