[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [idn] A question...

Soobok Lee <lsb@postel.co.kr> wrote:

> Not all chinese people are font/letter variant experts.
>
> Even for me, the subtle difference between U+6DF8 and U+6E05 would not
> be noticed without careful inspection.

For the purposes of my proposal, it doesn't matter whether you can see
the difference.  When you sit down at a keyboard with a domain name
in your mind and type it in, how many distinct strings have a good
chance of being produced?  There may be 2^n possibilities, but the
vast majority would require deliberate extra effort to type, and have
virtually no chance of being produced accidentally.  I'm hypothesizing
that usually only a very few of the many possible strings actually have
a good chance of getting typed in by a naive user.

Dan Ebert <dan@enic.cc> wrote:

> > For example, although there may be 8 different ways to represent
> > your school name, there are probably only two ways that are most
> > common/convenient (call them X and Y), and the other six ways are
> > much less common/convenient.  Suppose the registrar prohibited
> > anyone from registering those six ways?
>
> It may not be realistic to expect registrars to do this.  (Would each
> registrar have to check for the existence of a domain 'variation'
> every time someone wants to register a name?)

I am not proposing that the registrar allow an arbitrary pair of the
128 variants to be registered, whichever two are requested first.  I am
proposing that the registrar adopt an algorithmic policy for deciding
whether a given string is admissible (regardless of what else may or may
not be registered).  Once the algorithm has decided that the string is
admissible, the registrar would check whether it's already registered
(as usual).

If the requested string is inadmissible, the registrar would probably
offer some suggestions for "nearby" admissible unregistered strings,
just as registrars today offer suggestions when the requested string is
already registered.

> If this prohibition is to occur would it be better achived at the
> registry level?

Maybe.  In this discussion I have not been thinking about the
distinction between registrars and registries.

AMC