Re: [idn] IDN refocus, v4

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

James Seng/Personal wrote:

> applications will check validity of i18n domain names.
> policy will check validity of i18n host names.

Yes, more or less.

> 1. Do we need more than one CCS.
>    Conclusion: No, we only need one and ISO/IEC 10646.

Agree

> 2. Do we need more than one encoding (TES or CES) of ISO/IEC 10646?

We have more than one media (application) format. If we are to encapsulate
the domain names into the media efficiently, we need a TES for each of the
media formats.

I'm not sure we need to specify any CES rules at all. We only need to
specify the canonical CCS values, and do not need to specify a CES as far
as I can tell.

> 3. If answer to (2) is yes, then how many encodings cases do we need?

A trickier problem is finding the encapsulation substrate. Do we specify
encodings for protocols, or for data formats? In the latter case, do we
define TES for every registered MIME charset/encoding? I mean, do we
really want to define transfer encodings for ISO-2022-JP (including all of
the charset switching stuff)?

I would say that we should start with two core TES formats: 7-bit for
US-ASCII media, and 8-bit for UTF-8 media. We should also discuss the
issues surrounding the need for secondary transformations for all other
media formats and leave it up to the protocol and/or data working groups
for implementation. EG, if an email message contains ISO-2022-JP then our
suggestions should result in the 7-bit encoding being used, but we should
not have to specify that mapping explicitly.

> 3.1 For each cases, what is the right appropriate encoding?

If the approach above is the one that is taken, ACE and UTF-8.

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