Edmon Chung <[EMAIL PROTECTED]> wrote: > I think it makes a lot of sense to not force nameprep on the > application, but instead make it optional but mandatory on the server > end. Just like what we have nowadays for English.
I would have no objection to a newDNS protocol allowing unprepared non-ASCII text in queries and requiring the server to perform the preparation. (Well, I might wonder about the computational load on the server.) But I don't like the idea of unprepared ACE forms. It would mean disagreement about whether two ACE labels match. Someone applying the traditional ASCII comparison might say they don't match, while someone applying your proposed decode-prepare-compare method might say they do match. The conversions were designed so that comparing the ACE forms always yields the same answer as comparing the non-ASCII forms. The conversions are designed so that "unprepared ACE forms" don't exist; if you think you have one, you will find that it doesn't decode, it's just an ASCII label that happens to begin with the ACE prefix, but it's not an ACE. AMC
