Lower casing

[John C Klensin]

--On Saturday, January 29, 2011 9:15 PM +0000 Shawn Steele
<Shawn.Steele at microsoft.com> wrote:

> I'm somewhat concerned about the ensuing compatibility :)  It
> seems like vendors have 4+ options right now.
> 
> * IDNA2003
> * IDNA2008 (no mapping)
> * 5895 mappings
> * UTS#46 mappings
> 
> It is unfortunate that there's this much confusion.

Uh... Yes.

To make things worse, 

* ICANN is deploying (or letting others deploy, depending on how
you look at it) new TLDs that are required to be
IDNA2008-conforming.   If someone comes forward and says "we
need one interpretation of a string to be a variant of the
other" in the TLD name, they get one or both delayed, possibly
until DNSng.

* While we've been concerned about situations in which all name
resolution on a host goes through a single OS-based "resolve
name" interface, the IDNA model in which individual applications
do their own IDN (at least partially non-ASCII-string) to
A-label processing also exists out there.  In applications that
do their own IDN processing, there is the potential for a single
user sitting in front of a single computer to see different
behavior in different applications, or even in different
interfaces to the same application.

One way of looking at the problem with the four cases you list
above is that at least the advocates of each of the last three
believe that, if only everyone adopted their model, all would be
well.  And each of them is more or less correct; not that
knowing that does anyone much good. (I'm assuming that everyone
more or less agrees at this point that "IDNA2003 forever" would
not be a good idea.)

    john