procedure vs special form Marc Feeley 12 Mar 2001 17:15 UTC

The current proposal specifies that "error" is a procedure.  It could
instead specify that it is a special form.  The main advantage of a
special form is that it is possible to report the precise location
information (file, line, etc) in addition to the error message.
If "error" is a procedure, it is much harder to do so, in particular
when "error" is tail-called as in:

(define (f x)
  (if (< x 0)
      (error "x is negative")
      (+ 1 (sqrt x))))

The continuation that "error" will receive does not point back to "f",
so I don't see how "error" could easily report that the error occurred
in "f".

Marc