When/how mapping happens

[Harald Tveit Alvestrand]

Shawn Steele (???) skrev:
> Vince mentioned:
>   
>> The current protocol spec has an appendix in which a two-lookup method
>> is suggested. If an non-mapping look up fails, then the IDNA2003-like
>> mapping function would be applied and another look up attempted. there
>> are other possible implementations, of course.
>>     
>
> I would prefer that mapping be applied first (if desired), then lookup. 
In terms of the algorithms specified in draft-ietf-idnabis-protocol-11 
appendix A:

The A.1.1 and A.1.2 algorithms will give two lookups in the case where 
IDNA2003 and IDNA2008 results in different Punycode strings (I think the 
note at the end of A.1.1 is in error).

This will (as far as I understand) only occur in the case of our "four 
problem characters"; for all other strings, both algorithms give at most 
one DNS lookup.

A significant issue is whether the mapping should be exactly the 
IDNA2003 ToASCII function or some other mapping; applying the IDNA2003 
mapping means that we cannot reach labels containing our "four problem 
characters", so if one applies mapping before lookup, one has also 
assumed that the mapping to be applied is different from the IDNA2003 
ToASCII algorithm.

                  Harald

>  My reasoning is:
>
> * There should be no mapping that maps a valid label to a different label. (In other words, my suggestion should have no net impact)
> * If mapping is permitted to happen first, then applications can get away with one lookup, otherwise they have to try two lookups, which is harder at some layers.
>
> * And, of course :), I'd like to reduce the steps in my implementation as much as possible, so this would help me.
>
> I do not mean to give any more weight to the mapping step by this suggestion, I just hope to streamline implimentations.
>
> - Shawn
>
>  
> http://blogs.msdn.com/shawnste
> _______________________________________________
> Idna-update mailing list
> Idna-update at alvestrand.no
> http://www.alvestrand.no/mailman/listinfo/idna-update
>