Re: [idn] Requirements I-D

[Patrik Fältström <paf@swip.net>]

At 16.28 +0100 00-05-15, RJ Atkinson wrote:
>At 20:59 15-05-00 , A. Vine wrote:
>
>>Maybe but how do you fold A? to a? or a with grave accent? Which is 
>>the match?
>>I don't know the anwer to that, I'm just posing a question.
>
>One way is to do this in the manner specified by another group.  For example,
>UNICODE has a standard set of procedures for canonicalisation that would
>cover this situation.  My point was simply that we need a single method,
>independent of the details of that method, for canonicalisation of DNS names.

Exactly, what you, Andrea, start talking about is what normally is 
called canonicalisation and not case folding, at least in my 
terminology.

To take a step back:

Can we at agree that we need at a maximum one predefined method which 
handles case folding/canonicalisation and such? I.e. as Ron states, 
it is needed for global interoperability to know beforehand what will 
be unique DNS names and what is not. Uniqueness is locale dependent 
BUT we can still define for IDN a specific rule, for example as Ron 
states, point at some other organization which happen to have some 
algorithms already. Regardless of the algorithms correctness, it is 
_one_ algorithm which is what we need for IDN.

Note that I am saying "at a maximum one", which means that we might 
not have this at all.

     Patrik