Re: a (revised) draft of R5RS

[Gerald Jay Sussman <gjs@martigny.ai.mit.edu>]

I am afraid that I have to agree with KMP here, though I wish he would
not talk in implementation terms involving such extraneous concepts as
stack groups.

Both UNWIND-PROTECT and DYNAMIC-WIND have shown themselves to be
essential in practical programs, for different reasons.  They
implement capabilities that are not possible to obtain by simple
combinations of other mechanisms.

The reason for these mechanisms is that we often must deal with
systems that have state variables that we must manipulate.  It is
possible that we could make a language construct for each of them that
does it "right" and only advertise that method, but the mechanisms of
UNWIND-PROTECT and DYNAMIC-WIND allows us to experiment with and build
the features we want within Scheme, without having to resort to some
other "implementation" language.