Taylor R Campbell scripsit:

> I suggest providing another procedure REFERENCE-BARRIER that
> guarantees whatever object you pass to it will not have been GC'd
> before the call to REFERENCE-BARRIER.

How is that different from stashing the object in a variable?

> The example implementation in Racket is incorrect.

Thanks, of course you're right.  I'm not really used to ephemerons yet.

> Some of the text was lifted verbatim from the MIT Scheme reference
> manual, including most of the paragraph starting `This procedure must
> be used with care...'.  This should be acknowledged.  (It's also not
> clear which procedure `this procedure' refers to.)

I'll improve the reference as you noted in your second email.

> I encourage providing some automatic tests.  Here's a place you might
> start:
>
> http://git.savannah.gnu.org/cgit/mit-scheme.git/plain/tests/runtime/test-ephemeron.scm

Good idea.

> http://mumble.net/~campbell/tmp/ephemeron.scm
> http://mumble.net/~campbell/tmp/ephemeron-hash.scm
>
> I would be happy to provide them under an appropriate licence for
> inclusion in the SRFI if you like.

Please do.

> Bruno Haible, `Weak References: Data Types and Implementation', 2005-04-24.
> http://www.haible.de/bruno/papers/cs/weak/WeakDatastructures-writeup.html

Will do.

--
John Cowan          http://www.ccil.org/~cowan        xxxxxx@ccil.org
You know, you haven't stopped talking since I came here. You must
have been vaccinated with a phonograph needle.
        --Rufus T. Firefly