Punycode & IMA/EAI

John C Klensin klensin at jck.com
Thu May 22 15:05:59 CEST 2008

--On Thursday, 22 May, 2008 21:46 +0900 Felix Sasaki
<fsasaki at w3.org> wrote:

> Am I right in the assumption that the *if* there is need for
> Powder to have 100% reversibility of mapping U-label <>
> A-label, this could be guaranteed with IDNbis as currently
> planned, but not with IDN2003?

Mostly yes.   For IDNA2008, one would simply have to enforce a
"no mapping" rule in the application (external to IDNA itself).
For IDNA2003, since the non-reversible case and compatibility
mappings are integral to the protocol, one can, strictly
speaking, not do that.

_However_ note that enforcing a "no mapping" rule in the
environment surrounding IDNA2008 invocations gives up some
compatibility with expectations for IDNA2003 strings.  We have
some statistics (posted by Mark to this list earlier, I think)
about how many existing names in references, as seen by Google,
are dependent on the IDNA2003 mappings.   We have no real
information about how strong the expectation of mapping is held
in the community (although there are many strong opinions).

Conversely, one could effectively reverse-engineer IDNA2003,
applying the Nameprep mappings as a canonicalization method
before applying IDNA.  If one did that, IDNA2003 would be, for
this purpose, roughly equivalent to IDNA2008.  But one would
need to be sure that any domain names used in or with Powder
were in that reduced form, rather than the more permissive forms
permitted by IDNA2003.


More information about the Idna-update mailing list