[R6RS] Multiple-value unspecifiedness

[William D Clinger]

Mike wrote:
> Does the unspecifiedness of the multiple-value semantics extend to
> allowing crashes?  (In safe mode, of course.)  If so, what
> restrictions do we impose?

This issue is similar to the issue we faced earlier with
respect to the use of set-car! or set-cdr! on immutable
lists.

> Could somebody suggest some wording to go in the report?

If no values or more than one value are returned to a
continuation that was not created by call-with-values
and was not created by a begin expression, then an
exception might be raised.

We could haggle over whether the word "might" should
be "may" or "should", but the R6RS has to deal with
this kind of issue in several places, and should use
the same kind of language throughout.  For suggested
definitions of "must raise", "should raise", "may
raise", and "might raise", see draft/safety/safety.txt.

Will