Re: [idn] Prohibit CDN code points

["Masahiro Sekiguchi" <seki@jp.fujitsu.com>]

> It is possible if  we assume the ccTLD implied the language used .  So, they
> can be treated with different way based on the ccTLD name implied language.

No, assuming _language_ by ccTLD is a bad idea.

We only need to assume each ccTLD implies a well-defined
set of rules of mapping/matching domain names.

One of the difficulty IDN WG is facing is one universal
set of such rules appropriate all-over-the-world.  By reducing
applicable area of each set of rules smaller, it will be far
more easier to find the rules.

My scenario is as follows:

(1) IDN WG defines a basic set of rules with several options.
    One of the options can be: enable/disable SC/TC mapping,
    for example.

(2) Each TLD registry defines and propergates a concrete
    set of selections among those options, to be used
    for the domain names under the TLD.

    For example, .cn domain can say "SC/TC mapping must be 
    enabled for .cn" if it considers that decision serves the
    .cn domain name users best.  If another domain, say .us,
    finds contrary, it can say "SC/TC must be disabled for .us".

(3) The set of selected options for each TLD must be somehow
    notified to all DNS clients.  IDN WG must develop a
    reasonable way to allow it.  One possibility is to use
    DNS query against the TLD for a special RR.  (This is
    just an example idea.)


(4) DNS clients implements NAMEPREP module as a "parameterized"
    algorithm.  The parameters here is the set of selected
    options.

Note that the intention of this optional (or parameterized)
approach is to make negotiation/compromise easier.  If a ccTLD
considers almost all users of that particular ccTLD uses
same language, the ccTLD can use a set of options optimized for
taht language.  If it finds its users use relatively small 
number of languages, the ccTLD can compromise to find a set 
of rules suitable for those small number of languages.  In case
a ccTLD finds its user communities are so broad, it can choose
a set of neutral options that is *not* best for anybody but is
not harmful either.

The point is: assuming "a ccTLD is tied to a language" is
infeasible, but "to a set of processing rules" looks OK.