how did the idna theory start?

John C Klensin klensin at
Sun Jul 1 18:42:30 CEST 2012

--On Sunday, July 01, 2012 12:31 +0000 "Abdulrahman I. ALGhadir"
<aghadir at> wrote:

> The second point makes sense.

Two observations:

(1) The first point makes sense unless we were all willing to
deal with a delay of possibly decades.  New ways of doing things
in the DNS to which actual resolvers and servers have to adapt
take a _very_ long time to deploy, with no workarounds until
they do.   It has taken IDNA much longer to deploy than anyone
expected then, but the workarounds exist and it is possible to
adapt on an application-by-application basis.  There are
problems even then; see RFC 6055.

(2) While Patrik is exactly correct about the second, if you (or
others) want to ask this question, I think there is something
else that is worth repeating.   I believe that, although we
reached the right conclusion,  we underestimated, and often
continue to underestimate, the complexity of comparisons of
Unicode strings.   It is really tempting and easy to say "just
use Unicode" or even "just use UTF-8", without appreciating
that, as soon as the requirements involve issues more complex
than simple display, meeting expectations about how and why
strings should match get extremely complicated.  IMO, IDNA goes
about as far as is possible and gets it right within the limits
of the DNS, but lots of people don't want to accept those limits
(or are unable to understand them), leading to interminable
discussions about character mapping (not just as a forward
compatibility issue), "variants", various type of synonyms, the
appropriateness of transliterations and translations of other
strings, etc.


> -----Original Message-----
> From: idna-update-bounces at
> [mailto:idna-update-bounces at] On Behalf Of Patrik
> F?ltstr?m Sent: 1/Jul/2012 9:21 AM
> To: Abdulrahman I. ALGhadir
> Cc: idna-update at
> Subject: Re: how did the idna theory start?
> On 1 jul 2012, at 07:43, Abdulrahman I. ALGhadir wrote:
>> What is the main reason for not supporting the UNICODE in the
>> DNS protocol and to not use the hack-and-slash current way to
>> solve this issue?
> Because of mainly two reasons:
> 1. Backward compatibility with ascii domain names where domain
> names are in use in protocols.
> 2. The problematic part of IDNA is not the encoding, but the
> choice what characters to allow etc. If we one day need a
> completely different version of IDNA, then we can use a
> different prefix than "xn--" and still we have not destroyed
> the complete Unicode name space.
>    Patrik
> _______________________________________________
> Idna-update mailing list
> Idna-update at
> _______________________________________________
> Idna-update mailing list
> Idna-update at

More information about the Idna-update mailing list