Comments on draft-ietf-idnabis-defs-10

Andrew Sullivan ajs at shinkuro.com
Mon Aug 31 16:37:57 CEST 2009


Dear colleagues,

First, I wish to apologise for having missed this wrinkle when I did
my review last week.  I did in fact review Punycode in order to check
my assumptions, but I think my assumptions were too deeply-assumed, so
I completely missed this possible problem when trying out different
formalisations.  That said, having read the thread I'm not yet sure
there's a problem here.

On Mon, Aug 31, 2009 at 10:31:09PM +1000, Wil Tan wrote:
> "Fältström" is an invalid U-label because of the character "F". According to
> idnabis-defs-10, this makes its punycode encoding "xn--Fltstrm-5wa1o" an
> invalid A-label. Though invalid, it could still be looked up if the mapping
> step is employed (because idnabis-mappings-03 lowercases everything) or if
> the application chooses not to attempt to decode the A-label input to
> U-label.

Since mappings lowercases everything, then the invalid label is looked
up in the IDNA sense, but the resulting invalid A-label is not
resolved; instead, another label is resolved.

The big problem, then, is that on some systems "Fältström" appears to
get looked up, and on others, it may just fail.  This is the sort of
inconsistency that mappings is supposed to solve, but we know that
there will be such inconsistencies anyway: that's the cost of allowing
localization as opposed to trying exclusively for
internationalization.

Perhaps what is needed in the mappings document is a suggestion that,
if the input label is in fact mapped, the _mapped_ version of the
input label is what is returned after lookup?  This would cause the
input "Fältström" to be returned after lookup as "fältström".  (I
don't know if I like this idea, but I thought I'd suggest it anyway.)

A

-- 
Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.


More information about the Idna-update mailing list