New versions of IDNA documents

Now also ICANN in a posting. Discussions about the proposed spec has been whether the normalization (conversion) that was part of the IDNA we use today should be included this time as well or not. We who work with IDNAbis have made the proposal that conversion is not part of the standard. It is explained in detail in the documents themselves, but the short story is that conversion is very often context dependent (what language do you use, what other locale mechanisms do you like etc), while DNS is context free. Conversions have to be exactly the same for everyone. On top of that, it might be that registries at the time of registration want to use different equivalence lists between codepoints for registration / protection purposes, and we did not find it being useful to enforce specific mappings for those either. Many other arguments exists as well.

Instead we found it being important to say very specifically what codepoints in Unicode can be part of an internationalized domain name, and what codepoints can not. In the old version one had sort of three categories: Those codepoints that where ok, those that where not, and those that where mapped to the first group. It was when talking about IDN hard to explain which one of the three groups one talked about. Just the first, or the first and the last?