Re: Unified name properties

[Jim Blandy]

Keisuke's original suggestion --- that we simply provide a uniform
interface for finding the name of an object --- seems like a good
idea, and non-controversial.  I hope that doesn't get lost in the
discussion of what names mean, how they get set / computed, etc.

It's worth keeping in mind that object names are simply an
"environmental aid".  They're used by humans, not by programs, so:
- a certain amount of fuzziness is acceptable, and 
- we can change our system without worrying about backward
  compatibility.  We can do something quick and dirty, experiment with
  more elaborate ideas (like searching environments), and revert
  implementations with unpleasant consequences, pretty much at whim.

I did a (terrible) Scheme byte-compiler once that had all kinds of
ways to guess names for procedures.  If your code looked like this,

    (define adder
      (lambda (x)
        (let ((foobie (lambda (y) (+ x y))))
          foobie))

it would print the result of evaluating (adder 4) as

    #<procedure foobie in adder>

Certainly, the process was a heuristic one --- it's not hard to
construct code where it's not at all clear what name should be
assigned to the procedure.  But it was still helpful.

So I think people should concentrate on doing something helpful for
the user to see, and not worry so much about whether it has
well-defined semantics.