I want to point out that nobody should be using non-validating
UTF-8 converters; not only are they not conformant to the
Unicode standard, they also represent a security risk.
Mark, I clearly agree with this, and with Paul's point that
UTF-8 is not part of IDNA. But that, it seems to me, is an
argument for Simon's tests containing those tests, rather than
against it: if non-validating/ non-conforming UTF-8 converters
are out there, then testbeds that catch them would seem to be in
everyone's interest.