Re: [idn] Why follow IDNA with UTF-8?

[Paul Hoffman / IMC <phoffman@imc.org>]

At 12:56 PM -0500 7/15/01, Eric A. Hall wrote:
>I do not assume that I can predict all possible uses of inline UTF8 data.

Nor were you ever asked to.

>If you are only wanting one possible example, think of a protocol for
>replicating mail-routing maps.

That's rather obscure, but let's go with it.

>  Wouldn't it be beneficial for the admins of
>the affected mail servers to have a common format for that data?

Yes.

>  Wouldn't
>it make sense to use UTF8 for the IDNs in the router maps, given its
>usability with multiple languages?

Not necessarily. The format of the router maps could be plain text 
with UTF-8, plain text with ACE, ASN.1, and so on. You are mixing up 
document formats with DNS protocol.

>  Furthermore, wouldn't it be beneficial
>if the data and the IDN protocol both used the same encodings?

Probably, and that would be ACE.

>  How would
>it benefit anybody to manage ACE-encoded lists of mail domains?

Because ACE is being used as the host names that are being routed. 
You can't say "UTF-8 is needed for this protocol" unless UTF-8 is 
being used in the DNS.

>  How would
>it benefit anybody to replicate a UTF8 list that had to be converted into
>ACE for pre-processing?

The list does not have to be UTF-8: it could be in a wide variety of formats.

>  If UTF8 were available for the list contents and
>the pre-processing, wouldn't it that be beneficial to the users?

No more so than ACE, or ASN.1, or whatever.

>Clearly there are natural benefits to using UTF8 everywhere.

Sure, if you ignore most existing protocols and the severe problems 
during the wholesale upgrade of DNS servers. Have you now shifted 
from "ACE then UTF-8" to "never ACE"?

>  Providing
>access to UTF8 (whether with this mechanism or some other) allows for
>greater levels of functionality, servicability and usability.

You still haven't shown that. You gave one example in which a 
document format might be in UTF-8, but could just as easily been in 
any other format.

Do you have other, hopefully better, examples?
>  > >  New protocols are already required to use ISO-10646 for charsets,
>>
>>  That statement is incorrect. BCP 18 says ...
>
>I was referring to RFC2130.

RFC 2130 is not a "decree": it is an Informational RFC that is simply 
a report from a workshop. It does not "require" anything, and it is 
incorrect to say that it does. BCP 18, which came out after RFC 2130, 
is a synthesis of the ideas from the workshop. BCP 18 has 
requirements, but they do not apply to IDNA because IDNA does not 
deal with text content.

--Paul Hoffman, Director
--Internet Mail Consortium