The discussion of DNS wildcards has convinced me that we really need to
make use of the version field a MUST in the policy record.
While I agree that DKIM itself does not need this field, and have argued
for its supression as unnecessary in the past we do as a practical
matter require a clear mechanism for distinguishing between DKIM records
and SPF records. We also need to anticipate possible future SPF like
protocols.
I still think that the version field is unnecessary in the signature
header. If we need to make an incompatible change we should introduce a
new header value. RFC822 does all we need.