It Seemed Like A Good Idea, At The Time

It would appear that point #4 of my earlier proposal is entirely unnecessary.   A modifiable load function, like the one found in CHICKEN Scheme, can be implemented as follows:

(define (load path . evalproc)
 (let ((eval (lambda (expr) (eval expr (interaction-environment)))))
   (if (and evalproc              ;; Do some checks on the procedure
            (list? evalproc)
            (> (length evalproc) 0)
            (procedure? (car evalproc)))
        (set! eval (car evalproc)) ;; True
        #f)                        ;; False
   (letrec ((eval-datum (lambda ()
                          (let ((datum (read)))
                            (if (eof-object? datum)
                                path
                                (begin
                                  (eval datum)
                                  (eval-datum)))))))
     (with-input-from-file path eval-datum))))

Some of this code is probably unnecessary, but I felt it better to err on the side of checking too much before using the optional evaluation procedure.  I could also trivially rewrite it to use the more clear DO form, rather than a tail-recursive lambda.  Also, it might be more adviseable to use the standard LOAD function if no special evaluation procedure is given.  I’m torn on that last one.  Whatever.

But anyway, I see now that I really should have given the matter some deeper thought before immediately writing a lengthy proposal to change fundamental language features.  Over the next few days, I intend to work on a Scheme implementation detection library.  If I am successful, the two most important points of my earlier proposal will have been fulfilled, and I will be content.

It’s been a learning experience.