Re: [idn] additional comments from draft-ietf-idnra-00.txt

[Patrik Fältström <paf@cisco.com>]

At 08.50 +0200 00-08-29, Dan Oscarsson wrote:
>  >Because we don't want to have UTF-8 in names in application protcols
>>which is not defined to use 8bit characters.
>
>But we want all protocols to go 8bit! (and most are).

Of course, but they can not handle 8 bit in the addressing.

>  >A preliminary discussion with application area working group chairs
>>said that "domainnames in protocols should stay at 7bit, or go to
>>UTF-8" and further that "SMTP can be changed to handshake to UTF-8,
>>if fallback encoding exists and is well-defined, but HTTP will not be
>>fun due to the definition of URIs".
>
>Yes, SMTP can be fixed easily.
>HTTP is already 8bit. There is no problem to use UTF-8 in URIs.
>Except for the stubborn group who think an URI can only be ASCII.
>I have used 8-bits in URLs for many years.

What is needed is to define what the %-encoding of 8bit characters in 
URIs should be. See the draft which Larry has written.

>It is really high time that the ASCII only people let go of their idea
>that ASCII is everything. Most protocols can use 8 bits, most can
>easily use UTF-8 (which is not that efficient but we have it for the
>ASCII people. If will not help me who require ISO 8859-1 compatibility).

Dan, we had this discussion in 1992 aswell when we both worked with 
MIME...and I will not bring up the old arguments about "yes, 8bit 
might work, BUT the corners of the usage of the protocol is undefined 
as 8bit is not well defined, but work -- sort of".

We have different opinion on this issue, so I will not try to force 
you to think and belive in what I do belive in. :-)


The point with my message was that the big task will be to review the 
protocols which use domainnames to identify:

- If non-ascii usage is defined (charset definition etc)
- If non-ascii useage works
- If non-ascii usage can be fixed in the protocol specs
- If applications can be fixed to non-ascii usage works

Some applications are pretty easy (SMTP iff UTF-8 is chosen as 
encoding, and an ascii fallback is defined). Some are worse (HTTP 
because of the lack of proper definitions of some things). Some are 
really in bad shape (any usage of X.509 certificates).

I am, as I said before, ready to start work in Apps Area given some 
more constructive ideas from the IDN wg on what the choice should be 
in IDN solution.

    paf