Harald Alvestrand harald at
Tue Dec 1 11:17:41 CET 2009

Shawn Steele wrote:
>> I don't want anything, anywhere, to suggest that I have registered three 
>> domains, one of which is "paß, another one is 
>> "" and a third one is "".
> Assuming the rules are applied consistently, then they would be clearly understood by all.  If ß is mapped to ss as in 2003, and if we continue having case mapping for ASCII, then it is clear that you have only one registered domain.
> Should the rules change so that ß no longer maps to ss, then it is no longer clear.  Depending on the browser and usage it might appear that you have 1 or 2 domains.  That's part of why the rules need to be stable and applied consistently.
well.... "clearly understood" is clearly (sic!) not something that's 
readily available.

What I want, long term, is that if someone looks up "paß", 
they get an NXDOMAIN.

I know that I can't get what I want currently, because of what I regard 
as a technical mistake made in the design of IDNA2003. I would prefer 
the world to change so that I can get what I want, eventually, but know 
that it's unlikely to be completely consistently available for a number 
of years. (if complete consistency ever happens).

As for the size of the backwards compatibility problem:

As of Oct 13, Erik van der Poel said that the number of references with 
ß was:

0000DF  ß       0.00001%

If we pick a random number and say that "all" represents 1 billion 
references, then this is ~100 references currently using ß. (Note: I 
have no idea of the base number. I could easily be off by orders of 

(The data at, 
referenced by, 
says 0.00022% - under the same assumption, this is ~2200 references. The 
problem's going away.)

But if I'm right, and if we could have one of those references changed 
to "ss" for every message sent to the mailing list about the eszett 
backwards compatibility problem, there wouldn't be a compatibility 
problem any more.


More information about the Idna-update mailing list