referencing IDNA2008 (and IDNA2003?)

John C Klensin klensin at
Fri Oct 22 05:02:02 CEST 2010

--On Thursday, October 21, 2010 22:36 -0400 Patrik Fältström
<patrik at> wrote:

> I completely agree, and fwiw people not really into the
> details of IDNA2003/IDNA2008 details should note what people
> sort of have said between the lines, with IDNA2008, it does
> not really matter whether A-label or U-label is in use, while
> in IDNA2003 you really wanted to use what in IDNA2008 is
> called A-label.
> And the incompatibilities when you look at this is really only
> in two codepoints, final sigma and the sharp-s.

Even then, if the A-labels are used exclusively, there is no
incompatibility.  If the A-labels were generated from IDNA2003,
those code points cannot occur and only "ss" and "lower case
sigma" will be present (and generated if the A-labels are
converted to display for by either IDNA2003 or IDNA2008).  If
IDNA2008 was used, then those characters may appear but only if
that was what existed initially.  And the "just convert, don't
test" approach used in IDNA2003 should cause the right things to
happen even if final sigma and/or sharp-s are actually encoded
into the A-labels (presumably by IDNA2008) but IDNA2003 was used
to produce the display form.  Only if the IDNA2003
implementation tests the actual code points produced by the
ToUnicode conversion against Nameprep will the difference even
be noticed.

> Something that I personally think is a "bump in the road" that
> will be handled (and is handled) by registration policy, which
> also will go away faster in reality than what we might be
> nervous about in theory.

> My suggestion is because of that also something like "use
> A-labels when comparing and storing strings in cookies". Done.
> Do not invent your own algorithms.
> It is much easier to then iron out the potential issues as
> real issues as both the real issues _and_ mistakes in the
> design of the tailored algorithm. Which unfortunately most
> likely will be greater than zero. Regardless of how hard we
> work.



More information about the Idna-update mailing list