Re: Further thoughts on preferred printable representations...

Roy Badami writes:
> The prefered printable representation[1] of an e-mail address is a
> vital concept, as I argue in my previous message.

One of the big flaws in IDNA is its failure to recognize this basic fact.
As I wrote in http://cr.yp.to/djbdns/idn.html:

   The value of a character-set expansion comes entirely from the
   visibility of the additional characters to users. There is no point
   in merely expanding the set of bytes allowed inside the computer; the
   internationalized domain name {alpha}{beta}{gamma}.com must be
   displayed with Greek letters on a typical user's screen.

The same issue arises for any piece of text. There's nothing special
about domain names, or mailbox names, or blah.html, or chat usernames,
or whatever tomorrow's textual identifiers will be.

> It is what I print on my business card.

There's a difference between what should appear on a screen for users
to see, and what should appear in print for users to type. As I wrote
on the idn mailing list two and a half years ago:

   Operating systems are all going to support direct input of Unicode
   characters by number. The ISO standard method is Shift-Ctrl-222E for
   character 222E.
   
   People who put hard-to-recognize characters onto their business cards
   are going to add a line giving the Unicode numbers, perhaps in boxes:
   
      [email protected]            (with a contour-integral sign)
   
                 +----+
      postmaster@|222E|.cr.yp.to       (in smaller type)
                 +----+
   
   The second line won't distract the people in the intended audience who
   recognize the contour-integral sign and have, say, Alt-S configured to
   produce it.
   
   End of problem. Notice that this doesn't require any extra DNS names.
   The conversion of Shift-Ctrl-222E to \342\210\256 is isolated inside the
   keyboard interface; other software doesn't have to worry about it.

The ISO standard mentioned is ISO 14755. Notice that it works the same
way for all text; it isn't something that has to be reinvented for
domain names, mailbox names, etc.

---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago