language question re: IDNs

Eric Brunner-Williams ebw at
Wed Oct 7 16:04:35 CEST 2009

Agree. There's not a lot of excitement over the first 63 digits of pi as 
a label (it exists, pleasingly enough), with, or without, "s" and "5" 
variations, or "l" and "1" variations, or even 
off-by-an-arbitrary-inner-digit-transposition ... umm ... "variations".

I'd prefer not to see "homomorphism" used, as someone is liable to try 
and invent interesting algebraic structures of glyph ensembles over 
which to attempt to map, while preserving the structure of the glyph 
ensembles, elements of equivalence classes to possibly other elements of 
possibly other equivalence classes.

A "no {algrebra|DNSSEC} here" disclaimer would be sufficient.


Andrew Sullivan wrote:
> On Wed, Oct 07, 2009 at 02:51:41PM +0200, Cary Karp wrote:
>> That's a very reasonable distinction in itself and surely nice to have 
>> waiting to put forward in any context where it might be useful. Again FWIW 
>> -- the ICANN w.g. really is focusing on homoglyphic identity (without any 
>> intention of exporting the terminology), and IDNA is explicitly restricted 
>> to labels and not entire names.
> But I think Debbie has a point that the human issue, which is surely
> what the "homoglyphic identity" is supposed to solve, functions at the
> level of the label, and not the glyphs.
> To use an example she introduced, cl and d can be homomorphic.  The
> important thing is context.  The homomorphism is important for
> instance, due to the confusion that could be sown by someone phishing
> on the label 'malwareclean'.  It wouldn't matter in the case of the
> label '46ghcl55txx', because nobody is going to misread the latter
> label in a way that makes the confusion more pronounced (at least, I
> suspect that's right).
> A

More information about the Idna-update mailing list