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

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

At 11:36 AM -0500 7/15/01, Eric A. Hall wrote:
>Let me expand on this by saying that any application which supports
>version negotiation is a candidate for UTF8 migration.

Not unless there is a possible need for that migration. And, again, 
you have not shown even a single example of that possible need.

>For protocols that do not support in-line versioning, once a critical mass
>is achieved and sustained (say 10 or even 20 years from now), I expect
>that many implementations will default to UTF8, even if the protocol
>doesn't.

Ignoring backwards compatibility for no advantage is not a common 
trait among application writers.

>  > So, to be clearer: please list the advantages to the end user of the
>>  IETF creating a protocol that doesn't use the same DNS as the rest of
>>  the IETF protocols.
>
>It's not a different DNS. I don't know why you're saying that. The same
>authoritative servers and the same zone data is being used. What is
>different about it?

It uses a different part of the DNS protocol, namely the marking of 
the UTF-8 requests and responses. We are talking about *protocols* 
here, not file formats.

>The advantages may or may not be compelling at this point.

Well, you haven't given a single instance of one that might be 
compelling, so it is more accurate to simply say "there are no known 
advantages now".

>  It is most
>likely to simply be a matter of pragmatism when it happens (not "if" but
>"when").

So, although there is no need now, you firmly predict that there will 
be and therefore we have to act now with a protocol. How do you know 
that your protocol will in fact meet the future need? There are lots 
of different proposals for how to do marked UTF-8 in DNS; why are you 
so sure that your marking plan will work for the as-yet-unknown needs?

>  New protocols are already required to use ISO-10646 for charsets,

That statement is incorrect. BCP 18 says that protocols that use 
character data are required to at mark the character data with 
charsets, and at least be able to use the UTF-8 charset, but are 
explicitly allowed to use other charsets. IDNA does not use character 
data: it encodes labels.

>so how long will it be before the IAB simply decrees that UTF8 is the
>preferred encoding as well?

There are IAB members who are following this list; maybe you should 
let them answer the question instead of answering it for them.

>  The ISO-10646 decree certainly wasn't driven
>by user benefit, why would this one?

BCP 18 most certainly was driven by user benefit; to say otherwise is 
fairly insulting to the many people who worked on it. Mandating that 
any protocol that can accept charsets MUST accept UTF-8 greatly 
increases interoperability (which benefits end users) *and* makes 
sure that all applications use at least one charset that uses all of 
the scripts in the world (which benefits end users outside 
English-speaking countries).

Please re-read BCP 18 before you use it in the context of this discussion.

--Paul Hoffman, Director
--Internet Mail Consortium