No internal DEFINE-SYNTAX?

["R. Kent Dybvig" <dyb@garbo.cs.indiana.edu>]

> In the R4RS macro appendix it says:
> 
>   Furthermore, there is no DEFINE-SYNTAX analogue of the internal
>   definitions described in section 5.2.2.
> 
> Is this because it leads to an contradiction?  Couldn't internal
> DEFINE-SYNTAXs be transformed into a LETREC-SYNTAX?

It doesn't lead to any more semantic difficulties than internal
defines.  Furthermore, a group of internal define-syntax forms can be
converted into a letrec-syntax expression.  However, a group of
internal definitions including both "define" and "define-syntax" forms
cannot be similarly converted.  Consider the following definition of
factorial:

   (define factorial
      (lambda (n)
         (define f (lambda (x) (if (= x 0) 1 (* x (g x)))))
         (define-syntax g (syntax-rules () ((g y) (f (- y 1)))))
         (f 10)))

This cannot be converted into a letrec-syntax for g with a letrec for f
inside; nor can it be converted into a letrec for f with a
letrec-syntax for g inside.  In the former case, f would not be visible
within g, and in the latter, g would not be visible within f.