Per wrote:
  > The examples that follow, however, are all of the shape
  >  (set!  (procedure-name ...) ...)

  That *is* a procedure application.  I guess the correct R5RS terminology
  is that the first set! operand can be a "procedure call".

The question is not whether this is a procedure call (believe me I know some
Scheme) but which class of  "procedure calls" you want to admit. From the
other remarks in your response, it seems you don't mean procedure call at
all but something that looks like a procedure call but only for some small set
of procedures (car, cdr, string-ref).  At least you didn't bring address-of
calculations from C.

Also, there are Schemes out there that implement define-syntax. It seems to me
that what you want to add can be implemented in define-syntax, tested, and
posted
with a test suite. Others have done something like this.

;; ---

Would it be too much to ask for a R5RS-style specification of SRFI's that
extend/modify Scheme's core syntax and semantics?  I can see that SRFIs
that propose a collection of functions are specified by implementation  and
description.

I would like to appeal to the SRFI editors to impose a simple standard in this
regard.
If there were a possibility to submit a meta-SRFI, I would do so now.

-- Matthias