I'm with you both. I wanted the IDN working group to define a subset, but ultimately I was persuaded that it was not feasible for the IETF to reach consensus on such a subset, and that subsetting could and should be done on a per-registry basis, inside the registry rather than the applications. But the slash-homograph attack now makes that approach (in its pure form) appear hopeless.
Of course, there may be some good reason why this doesn't fly.
Gerv