Re: [I18nrp] Mappings for IDNA2008 ?

["Patrik Fältström " <[e-mail]>]

On 5 Feb 2019, at 18:09, Asmus Freytag wrote:

> OK, if you make the choice that all of this should happen, then what?
>
> You'll have to use some tables / algorithm to do the mapping. Even when locale dependent (notionally) you have 90%+ of all cases that are unchanged and some small subset would get adjusted (tailored).
>
> That makes providing generic mapping tables useful.

I can as well say it is completely useless for the cases where the mapping *IS* special.

>> One can not (IETF found when discussing IDNA2003) do everything “in one go”.
>>
>> 1. Mapping
>> 2. Validation
>>
> You've lost me here. I would need a bit more background to understand this statement.
>
> I assume that "mapping" is something you do when prepping a user-supplied string for lookup.

Yes, case folding (LOCALE-dependent), normalization and what not.

> But not sure what "validation" refers to.

Validation that a label really is an A-LABEL/U-LABEL, i.e. valid according to IDNA2008 (or not), including (when you do call some library) conversion between A-LABEL and U-LABEL.

I.e. this discussion comes up when programmers ask how they should "convert a random string to punycode".

Some people think you can call convertToPunycode($randomString) and what I say is that you have to call mapAndNormalize($randomString) and then toALabel($randomstring) and toULabel($randomString).

   Patrik