Stability of valid IDN labels

[Patrik Fältström]

On 22 apr 2008, at 20.33, Mark Davis wrote:

> Patrick, I don't understand your reasoning.
>
> 1. Why must we guarantee to developers that DISALLOWED never changes?

The other way around.

We tell developers what codepoints are disallowed. They add those  
codepoints to be banned in their user interface. If then later we  
allow registration, we will have tons of applications out there that  
can not be used to access domain names that include that codepoint.

> I agree that it is important for DISALLOWED to not have any but  
> exceptional
> changes, but why is it important for developers to be able to depend  
> on it
> never changing? They can't depend on various changes over time: they  
> can't
> depend on it not growing; they can't depend on UNASSIGNED not  
> shrinking; and
> so on.
>
> For example, as an exception, suppose that
> U+02DE<http://unicode.org/cldr/utility/character.jsp?a=02DE>( ˞ )
> MODIFIER LETTER RHOTIC HOOK were to to be added in some future version
> to the exception list, because it was found to be needed for a  
> particular
> African language. This would involve changing from DISALLOWED to  
> ALLOWED.
> What would be the practical problem that this causes?
>
> We at Google, and I'm sure others, have lots of code that depends on
> backward compatibility -- that once an identifier is valid, it stays  
> valid.
> That allows us to always update to the latest validity checks on all
> identifiers, whether they are currently found or stored in  
> databases. (A
> one-time hit is containable for IDNA2003, but ongoing changes would be
> untenable.)
>
> But anyone who stays current with the IDNA2003 specs always be  
> expanding the
> valid identifiers.

People use software without upgrading until their computers stop  
working. And that is 10+ years.

> 2. How can such a promise be guaranteed? What is to prevent IDNA2012  
> from
> obsoleting IDNA200x and moving one character from DISALLOWED to  
> PVALID?

Absolutely nothing of course. But when that choice is made, the people  
that make that decision must know that there will be software out  
there deployed until maybe 2022 or even longer that will not allow  
that codepoint in domain names.

    Patrik

> On Tue, Apr 22, 2008 at 12:38 AM, Patrik Fältström  
> <patrik at frobbit.se>
> wrote:
>
>> On 22 apr 2008, at 03.49, Martin Duerst wrote:
>>
>> What may happen
>>> is that DISALLOWED is treated more closely to the way it is in
>>> IDNA2003: okay to query, but forbidden to register.
>>>
>>
>> I object to any kind of idea like this.
>>
>> This makes it impossible for application developers to filter, and  
>> there
>> is no way it is possible to control "registration" of DISALLOWED  
>> codepoints,
>> and the latter is the reason why application developers have to  
>> filter out
>> DISALLOWED codepoints completely.
>>
>> DISALLOWED is DISALLOWED. Forever. Period.
>>
>>  Patrik
>>
>>
>> _______________________________________________
>> Idna-update mailing list
>> Idna-update at alvestrand.no
>> http://www.alvestrand.no/mailman/listinfo/idna-update
>>
>
>
>
> -- 
> Mark