Re: [DNSOP] draft-liman-tld-names-04

[Masataka Ohta <[e-mail]>]

Some non-ASCII name wrote:

>>    As the upper case character corresponding to 'y' with diaeresis
>>    is not uniquely defined, for example, localized domain names
>>    are unusable
>>
>>    It should be noted that, extended case insensitivities beyond
>>    European characters, such as correspondence between Chinese
>>    ones, the problem is even more unsolvable.
>>
>> It should clarify how localized domain names are useless.
>
> I have two problems with this text. First that I think it is wrong.

Then, what is the upper case character corresponding to 'y' with 
diaeresis under localization context of ISO 8859/1, which is what
you are using in your mails as:

	From: =?iso-8859-1?Q?Patrik_F=E4ltstr=F6m?= <[e-mail]>

?

 > Secondly that the way IDN and other localization initiatives is
> by being a) backward compatible [so that the matching algorithm
 > can not be changed -- regardless of what you want] b) because of
> [a] it pushes the "case insensitivity" matching by asking
 > applications to canonicalize in a way so that if
> canonicalize(a) == canonicalize(b), then a == b, i.e. not do
 > the canonicalization in the DNS protocol itself. Because the
> matching function can not be changed.

If extended case insensitivities or canonicalization is not
handled by DNS severs and resolvers, which requires universal
definition, it causes unsolvable problems of:

	1) cache inefficiency

	2) wasted operational effort for synchronization of
	multiple zones with effectively same FQDN

	3) exponential explosions of the number of labels with
	effectively same FQDN

	4) doubly exponential explosions of the zones with
	effectively same FQDN for DNSSEC

That's why I wrote postscript in my previous mail.

PS> While it is trivially easy to define ASCII representation of
PS> Unicode, it is impossible to operate it, especially when
PS> (extended) case insensitivities are involved.

> and let the explanation on case insensitivities (etc) are in
 > the RFC589x IDNA2008 RFCs.

It's simply not operational.

						Masataka Ohta