[26562] in North American Network Operators' Group
Re: =?iso-8859-1?Q?Eeek_-_=2ENU_Domains_using_=D6=2C_=C4=2C_=C5=2C_=DC=2C_?= =?iso-8859-1?Q?=D1_-_And_More?=
daemon@ATHENA.MIT.EDU (Vadim Antonov)
Thu Jan 6 20:55:16 2000
Date: Thu, 6 Jan 2000 17:53:23 -0800
From: Vadim Antonov <avg@kotovnik.com>
Message-Id: <200001070153.RAA08665@kitty.kotovnik.com>
To: jmbrown@ihighway.net
Cc: nanog@merit.edu
Errors-To: owner-nanog-outgoing@merit.edu
"John M. Brown" <jmbrown@ihighway.net> wrote:
> Ahh, lets just move to UniCode! :) I mean 63 char SLD's and then this...
Ahh, just don't.
Domain names are specified as case-insensitive. You cannot convert
uppercase string in Unicode to lowercase (or vice versa) without
knowing about all collation rules in all possible languages.
(And, yes, precisely because of that your paranoid resolver won't be
able to compare results of a reverse DNS lookup with what A records say -
even if you don't plan to use domain names with Devanagari in it,
people from these domains may visit your server).
Having to install tons of locale configurations just to get name
resolver to work does not strike me as a good idea.
--vadim
PS Unicode is broken-as-designed - it requires rich text format
to do elementary things like lexicographic sorting or case-insensitive
comparisons. If you do have RTF with language tags of some kind,
Unicode is redundant.
