Issue: HASH-TABLE-GC (no proposal)

[David A. Moon <Moon@STONY-BROOK.SCRC.Symbolics.COM>]

    Date: Tue, 18 Oct 88 13:42 EDT
    From: Kent M Pitman <KMP@STONY-BROOK.SCRC.Symbolics.COM>

    There may be some little details to attend to, but abstractly the idea
    of a table with weak key links is both simple and powerful. I think we
    should look at the idea quite seriously. The question I have for those
    who think it's premature to do this is: what is the worst case scenario
    you can imagine for introducing this feature prematurely? 

If it's this particular feature, not too bad.  There seemed to be a good
deal of confusion or disagreement about exactly what feature we were
really talking about, how to define "weak" in a portable way, whether to
include finalization also, etc.  It's all that ill-definedness that
bothers me.  The standard might end up being defined in a way that is so
implementation dependent that it's very difficult for anyone but its
author to implement.  Alternatively, the standard might be so vague as
to be meaningless.

							      Are you
    envisioning a feature that is hard/expensive to implement? unreliable
    in some way I'm not thinking about?  turns out to not be something
    people want? Are the objections technical, economic, cultural ... or
    superstitious?

Perhaps my objections are more to the (envisioned) process than to the product.
If there was a concensus in the community as to what feature we were
talking about, and how to define it portably, then assuming it wasn't
unduly expensive to implement (which I think should not be a problem,
although I'm not familiar with the internals of many implementations)
and that it was not a burden for users (which ought to go without saying,
no incompatible change should be involved), then I wouldn't think it
was premature.  So maybe I'm worrying unduly.  But it's very late in
the X3J13 process.  The practical alternative is likely to be a few
Lisp vendors setting a de facto standard.