Re: [idn] lowercased hostnames (was: naming syntax rules)

["Eric A. Hall" <ehall@ehsco.com>]

Dan Oscarsson wrote:

> ACE is only how to encode non-ASCII into ASCII. It does not
> require things to be lower case.

In the context of this discussion it does. For systems that decode ACE
into the canonical UCS characters, the unencoded characters can be either
case. For systems that DO NOT decode ACE, the unencoded characters must be
case-preserved.

For example, consider a PTR which returns RR data that contains an ACE
encoded mixed-case IDN. If that RR data is received and processed by a
legacy system which does not decode the ACE data, then any subsequent
forward lookups must match the encoded representation exactly. This will
require that the forward-lookup A RR have the exact same owner name as the
PTR RR data. Furthermore, this must be used for ALL forward lookups,
unless every encoded version of every possible case combination is
delegated simultaneously.

> And what is nameprep? Of what I have seen nameprep is how to
> convert names into a format that is needed to make IDNA
> work for ACE encoded names.

Nameprep defines the characters which are legal, their normalized forms
(preventing un-normalized and normalized equivalents from being delegated
to two separate bodies), and will also need to have much of the rules
we've been discussing lately (such as no diacritical marks in the first
character position).

> I do not want to remove nice features of DNS just because some people
> do not want to upgrade their DNS servers (and those people will
> never get DNSSEC or other new features).

The opposite is also true: we should not make the entire system incredibly
complex just so we can have pretty domain names with pretty casing. These
are identifiers, not hairstyles. We should be favoring efficiency.

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/