I just realized it's possible for collations to return "undefined" even if all the input characters are valid. So I guess I'll have to redefine my request a bit. Change: 4.6. (b) The decoded text from (a) MUST be converted to the charset expected by the active comparator. to something like: (b) The decoded text from (a) MUST be converted to the charset expected by the active comparator. The server MAY/SHOULD handle failures (e.g. the text is in an unknown charset, contains a sequence which is not valid according in that charset, etc.) by replacing the failed part with a unicode "replacement character" #FFFD (or equivalent in other charsets).
Description: This is a digitally signed message part