can (values) be the (an) unspecified value?

["R. Kent Dybvig" <dyb@cs.indiana.edu>]

Some R4RS procedures and syntactic forms return an unspecified value,
e.g., set-car!, write, and set!.  Will it be/should it be acceptable
for an implementation to return zero values instead of "an" unspecified
value?  This would make some (strange) conforming R4RS/IEEE programs
nonconforming R5RS programs, e.g.:

   (let ((y 1)) (let ((x (set! y 2))) y))

I would prefer that, in spite of this possible inconvenience, we allow
implementations to return zero (or in fact any number of) values when
the current report says an unspecified value is returned.  In other
words, I'd like for R5RS to use the phrase "returns unspecified values"
or simply "returns unspecified" (with a note up front saying that
unspecified means zero or more unspecified values) in place of "returns
an unspecified value" whereever it appears.

Kent