If we implement utf-8 (variable width encoding), then many of the IBM tests will work. (Well, at least the 5 categorized as IBMTestsOnNumbers will work (I know because changing that test to use encoding="ascii" and they all run. But,... we don't want to change the IBM tests.)
Implementing utf-8 is a matter of implementing StringFixedLengthInBytesVariableWidthCharacters, and StringFixedLengthInVariableWidthCharacters.
These are variations on the theme established by StringFixedLengthInBytes, which is our current workhorse.
I suggest we fix this in s3 because I don't think it's terribly difficult, so the effort to accomplishment ratio is high