[r6rs-discuss] compilers, design philosophy

From: Thomas Lord <lord>
Date: Fri Mar 23 22:50:51 2007

This isn't directly responsive to previous threads but
it speaks to a few themes from a few threads.

Shiro made an example of the idea of reifying environments
as assoc lists: the report shouldn't do it, in his example,
because it would hork optimizing compilation.

I don't think assoc lists per se are the right interface
for first-class, mutable environments -- something
more abstract is called for -- but it would be nice for
some future version of the report to provide for them,
optimizing compilation be damned.

In general, what you can reasonably and practically
and usefully do in an interpreter is a superset of what
an optimizing compiler can do anything useful with.

The report's obligation is more indirect: it should only
ensure that one can rite separately compilable modules
that, with a simple static check, can be proved to use only
features that a reasonable optimizing compiler can do
something useful with.

-t
Received on Fri Mar 23 2007 - 23:00:53 UTC