Comments on the Unicode Codepoints and IDNA Internet-Draft

mail@edmon mail at
Thu Jul 31 13:22:46 CEST 2008

Also, as I understand it, this is only done at "registration" not at
And the result should be one that would reject a "registration" should
unstableness be identified.

So, general operations remain the same, the only suggested change is when a
particular CONTEXTJ/O char is encountered (on second thought maybe not even
all CONTEXTJ/O would apply, just certain blocks representing these combining
characters), it would trigger the unstableness rule check.


PS.  I wonder if we could even apply a similar approach for RTL chars...

> -----Original Message-----
> From: mail at edmon [mailto:mail at]
> The suggestion is to have the rule only for those codepoints listed as
> CONTEXTJ/O (i.e. to list combining chars as CONTEXTJ/O).
> Not for the rule to apply to all codepoints.
> Edmon
> > -----Original Message-----
> > From: Stephane Bortzmeyer [mailto:bortzmeyer at]
> > Sent: Thursday, July 31, 2008 6:40 PM
> > To: mail at edmon
> > Cc: idna-update at
> > Subject: Re: Comments on the Unicode Codepoints and IDNA Internet-Draft
> >
> > On Thu, Jul 31, 2008 at 08:12:51AM +0800,
> >  mail at edmon <mail at> wrote
> >  a message of 234 lines which said:
> >
> > > toNFKC(toCaseFolded(toNFKC(label))) != label
> >
> > This is interesting but it completely changes the tables, which are
> > currently per character, not per label. This would imply to change the
> > other rules (for instance, from "cp is not in Exceptions" to "no cp in
> > the label is in Exceptions").
> >
> > With the current tables, checking if a label is legal is almost
> > trivial (only composed of legal characters, with the special annoying
> > rule of context-allowed characters). With your proposal, checking if a
> > label is legal means running Unicode routines such as NFKC.

More information about the Idna-update mailing list