Re: [idn] canonicalization

[RJ Atkinson <rja@inet.org>]

At 12:19 27-05-00 , James Seng wrote:
>so is there any conclusion here?
>
>should we
>
>1. leave this out and treat that it is not a problem for anyone? or;

Doing this means that if I email you a specific URL, you might not see
the same web page that I do -- caused by subtle differences in our DNS
situations.  This will be very very confusing to users and is a really
bad outcome.

>2. do some minimual canonicalization and then hope for the best? or;

Has the same issue as proposal 1.

>3. some other suggestions?

I'd propose that we say that canonicalisation happens in the DNS resolver
client according to the applicable existing UNICODE specification, which
we then cite.  We should also explicitly say that the coded character set
used on the wire for IDN is only ISO-10646.  For historic non-IDN DNS,
the character set on the wire needs to remain the existing subset of ANSI X3.4.
This also means that DNS servers only see requests that are already in
canonical form.

To do otherwise can lead to the situation where a single specific URL
might go to different pieces of web content, depending on variables
not obvious to typical users.  ISO does not have a document we can cite
for canonicalisation.  IETF ought not define its own canonicalisation rules.

Ran
rja@inet.org