Re: [idn] UCS/UTF-8 is needed for interoperability

["Soobok Lee" <lsb@postel.co.kr>]

----- Original Message ----- 
From: "Dan Oscarsson" <Dan.Oscarsson@trab.se>
> We need ONE format for character data. 

I agree. Legacy coded identifiers cannot ensure stable IDN operations.

>UCS encoded as UTF-8 normalised
> form C is the best choice.

I take very conservative position from this point.

NFC is not mature. I have long lists of NFC corrections to be made.
only new normalization can solve it. "new bag for new wine".
If we think UTF8 solutions as long terms solutions, there is no 
reason why we stick with old NFC.

Moreover, we should sync IDN standardization with additions of all *MODERN* scripts 
which should also be debugged and stabilized.   after that, we could freeze the 
unicode version for IDN.

This may be bad news for some commercial driving forces of IDN, because that
means the lucrative IDN (+solutions) market is still far from them.

> 
> Locally in a system you will still use a local character set.
> But think how easy the code will be:
> - One library to handle the conversion between UCS and local charcter
> set

The conversion to utf8 should be made on direct user inputs and should also
give immediate/interactive responses regarding to any unsuccessful conversions if any.
That is crucial for identifier security and stability.

The DNS is the most crucial kernel component of the INTERNET *OPERATING SYSTEM*. That is why
most paranoid and conservative criticisms of IDN efforts are always justified and welcomed. :-)

Soobok Lee

>   that can be used by all programs.
> - Much simplified parsing, no embedded character set switches.
> 
>    Dan