On Wed, Oct 29, 2003 at 09:32:20PM -0800, Bradd W. Szonye wrote:
> Shiro Kawai wrote:
> > There's another point, though.  There are two camps in Scheme
> > implementations regarding how to give the "fallback" value, both have
> > their own ground.  If srfi-44 gives a convincing rationale for thunk
> > approach, and becomes widely spread, it may be possible that
> > eventually the convention of giving fallback value converges to thunk
> > approach.
>
> I've been thinking about this, and I'd rather raise an exception than
> provide a failure thunk. SRFI-34 defines exceptions, and SRFI-35 style
> conditions could provide information about the failure. It seems to me
> that a language with sophisticated support for continuations should take
> advantage of that in failure interfaces. It also simplifies call
> interfaces, since you don't ever need to distinguish, "Is this procedure
> a thunk or a collection datum?"

There are two problems with this.  First, exceptions would complicate
the code quite a lot.  Instead of:

(operator val (lambda () calculation ...))

We now have
(with-handler
  (lambda (error) calculation
  (lambda () (operator val)))

Additionally, exceptions have widely varying performance
characteristics, whereas all Schemes are written to optimize
application of functions.

	Scott