U-labels, NFC, and symmetry

Andrew Sullivan ajs at shinkuro.com
Fri Apr 15 17:50:28 CEST 2011

On Fri, Apr 15, 2011 at 07:19:34AM -0600, Peter Saint-Andre wrote:
> How is it that NFC meets the symmetry requirement, but NFD does not?

NFC meets the symmetry requirement because a string in NFD just isn't
a U-label, period.  It might be equivalent to a string in NFC that is
a U-label, but it is not itself a U-label.

This is really no different from the other mapping cases.  Given
locale L and a string in ISO8859-N, you might know how to map string
S(L,ISO8859-N) into a U-label, at which point you could start doing
work on it.  But IDNA2008 is silent on that.

Similarly, given unnormalized Unicode, you could run it through NFC
(and maybe some other stuff) and get it to be a U-label, but it isn't
a U-label before that.

If what you're saying is that you want a definition of D-compatible
U-label, I am not sure whether that is practical.


Andrew Sullivan
ajs at shinkuro.com
Shinkuro, Inc.

More information about the Idna-update mailing list