Sound classes differ from normal datatypes in so far as we may try to generate them, where a given sound is to complex by using a fall-back procedure. This procedure only makes sense for sound classes, so splitting of the sound class models as a third datatype seems to be in order, as it would facilitate representation as well (e.g., we do not need to list each sound class for a given grapheme in our graphemes/ view in CLLD).
Sound classes differ from normal datatypes in so far as we may try to generate them, where a given sound is to complex by using a fall-back procedure. This procedure only makes sense for sound classes, so splitting of the sound class models as a third datatype seems to be in order, as it would facilitate representation as well (e.g., we do not need to list each sound class for a given grapheme in our
graphemes/view in CLLD).