Philippe Verdy <verdy underscore p at wanadoo dot fr> wrote:Wrong. Non-standard normalization forms are useful too, and can even be safe if they preserve one of the two standard equivalences (canonical or compatibility).
I agree that non-standard normalization forms may have benefits, as in your Korean example.
I respectfully disagree that they should be used. IMHO, the potential for confusion and lack of interoperability is greater than the benefit.
There's no interoperability problems with non-standard normalization forms (unless a process interface absolutely requires it, which would be wrong as it should be allowed to accept any canonically equivalent string, such as one of the 4 standard normalization forms)..
But for internal processing, and even for process output, a non-standard normalization form is still a valid Unicode text, and thus should be interoperable (if it is not, blame the other "conforming" processes that reject these textsor process them incorrectly).
Non standard normalization forms are used everyday, notably within renderers, because the NFD form is not the easiest ordering for converting text to glyph strings; same thing for collation, where an intermediate non standard normalization form greatly helps reducing the size of collation tables and the complexity of the algorithm...
