Standardizing on IDNA 2003 in the URL Standard
John C Klensin
klensin at jck.com
Thu Jan 16 07:33:09 CET 2014
--On Wednesday, January 15, 2014 19:42 -0800 Paul Hoffman
<paul.hoffman at vpnc.org> wrote:
> On Jan 15, 2014, at 7:03 PM, Jiankang Yao <yaojk at cnnic.cn>
>> Aà.com has been registered.
> Verisign has some registration policy for some TLD. How does
> that affect the discussion of the URL standard? This thread
> seems to have started as a discussion of updating the URL
> standard, but has morphed badly.
It might be relevant because Anne seems to quite often argue
that existing practice has to be preserved forever if that were
what happened. But, to add a little bit Paul's comment and
those of Andrew and John L., you (Jiankang) can't infer that
"Aà.com" has actually been registered from the tool at
First, what happens when I put "Aà.com" into the tool, what I
get back is
Your search for Aà.com returns the below results:
Domain Name: AÀ.COM (XN--A-SFA.COM)
Registrar: INTERNET.BS CORP.
Now, converting the request you made into all upper-case in that
display is, IMO, stupid: it is an answer, valid or not, to a
question different from the one that was asked. Even IDNA2003
doesn't encourage doing that.
More important, you don't know what was "registered", all you
know is what is found in the database when you look up that
string. Consider a search for fuß.com: the response says:
Your search for fuß.com returns the below results:
Domain Name: FUSS.COM
I think that is ill-advised because I think they should have
responded to that lookup with a "no, but applications will map
it into 'FUSS.COM', which has the following record..." if they
believe IDNA2003 is universal or with "no, but some applications
may map it into 'FUSS.COM' so a registration of 'fuß.com' would
be blocked. The record for "FUSS.COM" is..." if they are
conforming to IDNA2008. For the latter case, regardless of what
one might think about mapping, IDNA2008 clearly allows a
registry policy that would block "fuß.com" if "fuss.com" is
What is actually "registered" in the DNS isn't, for the first
case, "XN--A-SFA.COM". First, some issues about
case-preservation notwithstanding, the label is "xn--a-sfa".
Under IDNA2003, mapping "xn--a-sfa" through TOASCII yields
exactly the same thing that mapping it to an A-label does under
IDNA2008: "aà". I hope it is obvious that neither RFC 5895
nor UTR 46 affects the mapping from the "xn--" form to native
characters; certainly both IDNA2003 and IDNA2008 discourage or
outright prohibit making any sort of upper case conversion in
that direction (for the reason Andrew first mentioned).
So, the Verisign search may be an application that maps before
doing the actual lookup or it may be non-conforming to IDNA2008.
In either event, its report of what is going on is misleading
and, as others have pointed out, what they are doing has nothing
to do with what IDNA2008 (or IDNA2003) might require.
More information about the Idna-update