I notice that when normalizing a Unicode string to NFKC form, superscript characters like ¹ (U+00B9), ² (U+00B2), ³ (U+00B3), etc are converted to the corresponding ASCII digit (ex. 1, 2, 3, etc).
Does anyone know the rationale for this behavior? It seems like it's losing information in the process. For example, a superscript number usually has some contextual meaning.