AW: Q2: What mapping function should be used in a revised IDNA2008 specification?

[Georg Ochsner]

> -----Ursprüngliche Nachricht-----
> Von: idna-update-bounces at alvestrand.no [mailto:idna-update-bounces at alvestrand.no] Im Auftrag > von Mark Davis
> Gesendet: Mittwoch, 1. April 2009 02:17

> I believe that the simplest approach for compatibility, and for
> implementability, is to use the one I gave in a previous email, as
> replacement text for 5.3. That is, use the same structure as IDNA2003:
> toNFKC(toCaseFold(toNFKC(x))), then remove all default-ignorable
> characters but the joiners.
> 
> As I said, it appears that given the current consensus, we don't need to
> specially except eszett and sigma, 

I don't agree that there is such a consensus.

Concerning your last email about DENIC's opinion. Marcos also stated very clearly that:

"a) In the pre 2003 era, we wanted eszett to be a separate IDN character, 
available for registration. Eszett and "ss" are just two different things.
b) We had to accept the roundtrip-case-conversion "rule" introduced with 
IDNA 2003, though it was contrary to our preference. We have now got used 
to live with it."


> but should we decide that we really
> need to preserve them, then we can do it in the following way.
> 1. Find all the maximal substrings that do not contain the exceptional
> characters.
> 2. Convert each of those substrings with the above mapping.
> 3. Apply toNFC to the result
> (This is a logical statement; the implementation can be optimized.)

Your idea shows, that it is easily possible to have a transition scenario for these characters. Also can the registries organize a grandfathering process for the registration of domain names with the new characters.


90 Mio. Germans and Austrians use the Eszett in everyday life. Eszett has its own key on the German keyboard. Domain names are normally typed in in lowercase. I think it's worth to have an "exception" algorithm within IDNA making Eszett work, as well as a grandfathering process (sunrise) for the registration. 

To continue to map Eszett to ss in my eyes means to pay more attention to "as little code as possible" and avoidance of having to organize a transition. Looking at all the great effort that is put into IDNA - even for languages and character sets that are not actively spoken any more - I cannot believe that going on with that mistake from 2003 can satisfy our ambitions!

Best regards
Georg