A modest proposal

What about using Herman's phenomenon as a high-error cutoff? That way
we get something we can rationally justify, and would only need a
rational justification for a badness exponent and cutoff to settle the
whole thing. One way to get such an exponent would be using the convex
hull. The cutoff justification could be moats, or how many items we
want on the list.

--- In tuning-math@yahoogroups.com, "Gene Ward Smith" <gwsmith@s...>
wrote:
> What about using Herman's phenomenon as a high-error cutoff?

Works for linears -- but for other kinds of temperament?