On Tuesday, July 08, 2003 5:14 PM, John Cowan <[EMAIL PROTECTED]> wrote: > Peter Kirk scripsit: > Such a character could only be encoded if it were put into the list > of composition exceptions, because it would upset the stability of > normalization.
Even if listed in the Canonical Composition Exclusion list, this would not work: this list only refers to characters that are canonically decomposable into a character pair, and that MUST be decomposed and MUST NOT be recomposed when creating *either* a NFC or NFD form. There's a requirement that if two string are canonically equivalent, they have identical NFC form *and* identical NFD form. The reason is that Unicode algorithms must produce identical results on NFC and NFD forms. Only the compatibility decompositions would work as expected, i.e. the NFKD decomposition of the "abnormal" sequence of vowels MUST be still given with two vowels in canonical order. meaning that the NFKD or NFKC transformation would swap the vowels to their canonical order. -- Philippe.
