Re: [idn] WG last call summary

["Adam M. Costello" <idn.amc+0@nicemice.net.RemoveThisWord>]

vint cerf <vinton.g.cerf@wcom.com> wrote:

> it occurs to me that TC/SC equivalences create a signficant burden on
> registration - will any registration of a sequence of such characters
> result in the implicit registration of a potentially large equivalence
> class of strings made up of substitutions of SC for TC or TC for SC
> in all possible combinations? What about the matching question? Will
> a canonical form be required that is in some sense the "generator" of
> the equivalence class of all variations in TC/SC spellings?

It's up to the registries how to handle this.  One approach is to
prohibit mixed TC/SC labels.  When someone tried to register such a
name, registrars could suggest the pure TC and SC equivalents.

If many registries adopt the same rules for which labels are allowed,
applications could assist users who type disallowed labels, offering
pure TC and SC suggestions.

Another approach would be to define equivalence classes and register
whole classes as a unit.  Implementation tricks could be played to use
computation in the DNS servers instead of large tables, if desired.
This would cause each host to have many names, which would help some
things and complicate others.

These two approaches could be combined:  You register a whole class (so
that no one else can use any name in the class), and then it's your
choice whether the DNS servers answer for all members of the class, or
only for a few select members of the class.  This might be the best of
both worlds.

And of course another approach would be to do nothing, and allow each
TC/SC variant to be registered independently.

AMC