xn--p1ai

[John C Klensin]

--On Thursday, February 24, 2011 09:31 -0500 Lyman Chapin
<lyman at interisle.net> wrote:

> Simon,
> 
> I think that many of us expect the case-folding behavior of
> IDNA2003 to be preserved by applications (esp. web browsers)
> at least during the transition to IDNA2008 (e.g., following
> UTS #46, http://unicode.org/reports/tr46).

And some of us believe that the recommendation of RFC 5895,
which include mapping to lower case (see its Section 2, item 1),
are more appropriate than UTS #46.  

With the hope that this will be taken as a brief summary of a
sincere difference in opinion rather than as the starting point
for another rehashing of old arguments, there are multiple
reasons for that preference, some of which are explained in the
introduction to RFC 5985.  The two key ones that are relevant to
this particular issue are, IMO,

-- There are only a small number of characters for which mapping
to lower case produces a different result than that produced by
the case-folding operation used in IDNA2003.  Where the two are
different, the lower case results are more obvious/ predictable
to non-experts than the case-folding ones.

-- For the four characters that UTS 46 identifies as
"deviation", there really is no transition strategy other than a
rolling flag day.  As long as UTS 46 transition processing is
applied, labels that IDNA2008 considers valid and distinct
remain hidden and unavailable.  Whether they become available in
the future is essentially a "flip of the switch" issue.  If
registries choose to treat strings containing the relevant
characters as "variants" (whatever that means) of each other and
register all of them with identical semantics (whatever that
means), having some of them hidden and inaccessible from some
software packages but not others makes no difference.  But the
intent of IDNA2008 was to make it possible for registries to
treat the strings as different, consistent with behavior in the
cognitively-relevant languages, if they considered that
appropriate.  UTS 46 "transitional" behavior essentially
overrides those decisions and will continue to override them as
long as it is applied.   Because the Internet continues to grow
as a high rate, some of us believe that the pain is best
incurred sooner (likely with fewer possibly-affected names) than
later (with lots more of them).   Others do not find that
position convincing.

    john