> ascii-string clearly has to accept any string at all, but I think
> that's a special case.  The other procedures should I think accept any
> character or any exact integer 0-127.

Coming from a country that uses an extended Latin alphabet, it's quite
weird if (ascii-letter? #\Ä) blows up. Of course no implementation would
have such a severe limit on the maximum processable integer.

Similarly, U+1D400 is Mathematical Bold Capital A. That's a letter so
it's a bit puzzling if you can't do (ascii-letter? ...) on it.

We could simply require integer arguments to be fixnums. If some Scheme
has such a constrained fixnum range that some of its characters are
bignums, then its users will just have to suffer when hitting the upper
end of that range :) If fixnums really are so constrained, users
probably have bigger problems than ASCII routines blowing up on arcane
inputs.

Anyway, doesn't Unicode allocation proceed quite slowly now that it's
gone past the Basic Multilingual Plane? Do some private use characters
have really big values?