[R6RS] Multiple-value unspecifiedness

[William D Clinger]

Mike wrote:
> I'm still unclear on whether we're allowing crashes in case no
> exception is raised.

Again, I think this is a larger issue.  What does it
mean to allow a crash?  Are crashes allowed even if
an exception is raised?  If an implementation
restriction is violated?  How do the default exception
handlers behave?  What is the meaning of safe mode?

There is also the larger meta-issue of why we might
want to answer these questions.  Are we trying to
describe behaviors on which programmers can rely?
Are we trying to proscribe implementation behaviors?
Is it just PR?

Whatever our answers, we need to keep edge cases in
mind: third-party libraries written in C or assembly
language that can do arbitrarily nasty things to our
guarantees, embedded systems where a clean exit might
as well be a crash, out-of-memory errors, et cetera.

Will