On Fri 20 May 2011 21:44, John Cowan <cowan@mercury.ccil.org> writes:

> Andy Wingo scripsit:
>
>> There is also the approach of using no type at all and returning the
>> values on the stack.
>
> Indeed, though that can lead to very strange behavior.  For example,
> if the identifier "map" is known to be bound to its R5RS definition, a
> compiler can reject (map (foo)) as erroneous, but under a "return on the
> stack" implementation, (foo) might return both the function and the list
> to map.  I think it was Henry Baker who pointed this out.

You are confusing the stack with the continuation of the evaluation of a
procedure's argument.  I was referring to the technique described here:

  http://www.cs.indiana.edu/~dyb/pubs/mrvs.pdf

Andy
--
http://wingolog.org/

_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports