Re: [idn] case preservation

["D. J. Bernstein" <djb@cr.yp.to>]

Patrik writes:
> characters which look "similar" like 'o' and 0, or '1' and 'l'

I am not talking about the momentary confusion between ao1.com and
aol.com. I am not talking about characters that merely look ``similar.'' 
I am talking about characters whose standard glyphs are IDENTICAL.

A properly typeset uppercase Alpha is IDENTICAL to a properly typeset
uppercase A, just as a properly typeset tab is IDENTICAL to a properly
typeset sequence of spaces. It is simply NOT POSSIBLE to tell the
difference between AOL.com and AOL.com.

This does not mean that Alpha and A are the same character, or that tab
and space are the same character. It does mean, however, that you've
done a half-assed job with the current version of nameprep.

If you're not going to do a thorough job of nameprep, then you shouldn't
be trying to deceive users into believing that IDNs can be safely typed
from textual printouts. It would be better to omit nameprep entirely;
then users will stick to mechanisms that work, such as clicking on URLs.
(Bad characters should still be avoided in registrations, of course.)

---Dan