Re: [idn] Requirements I-D

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

At 1:04 PM -0700 5/15/00, A. Vine wrote:
>There will have to be rules upon registry leaving accented 
>characters out and so
>on, or checking accented characters again unaccented characters, or 
>restricting
>input to force the user to understand how something is cased in DNS.

This is precisely why we now have a nice diagram in the requirements 
document. At every place that is between boxes, we can define what 
can/must/cannot/mustnot be passed in each direction. Personally, I 
believe it is wrong to force the user to do anything; all 
canonicalization changes should be done at other points in the chart.

>   The DNS
>may happily interpret any way it chooses, but it will have an effect 
>on the way
>the data is entered in the first place.

Quite right, just as it is now.

>   Perhaps these restrictions can be made
>clear to all users in all situations where they type in a domain name with
>characters which normally share an uppercase form with another lowercase
>letter.

It is no different than the kinds of restrictions we have now, other 
than it is not so onerous. No we say "if it isn't an ASCII character, 
forget it" and "if it is an uppercase ASCII character, it will be 
treated as identical to a lowercase ASCII character". These are rules 
that users have gotten used to.

>   This may just be a rendering difference if the data is properly
>canonicalized.  But it is something to look at and think about.  Perhaps it is
>necessary to be in the requirements.

Yes, that's why were' discussing it here! James wanted to drop [25] 
from the requirements document.

--Paul Hoffman, Director
--Internet Mail Consortium