[vip] Types of variants: do we have consensus?

[Andrew Sullivan]

On Mon, Jul 25, 2011 at 03:58:15PM -0700, Alexey Mykhaylov wrote:

> I would like to point out (considering that this group deals with IDN TLDs
> issues) that the FIRST requirement for IDN gTLD application in gTLD
> Application Guidebook states as the following: 

> 
> "Meaning or restatement of string in English. The applicant will provide a
> short description of what the string would mean or represent in English." 

Thank you for bringing this requirement to my attention; I somehow
missed it in previous readings of the guidebook.  I'm sure you can
work out what my (personal) opinion of this requirement is.

> I think it is safe to claim that TLDs do have meaning _associated to them_

Semioticians will tell us that _everything_ has meaning associated to
it.  Of course DNS labels have more or less meaning for a given
person, and over time a user community might come to converge on a
conventional meaning.  On the other hand, I've often heard it said
that .org domains are for non-profits.  But there are no restrictions
on the non-profit status of registrants using a .org address.  And
indeed, as far as I know the owners of slashdot.org hope to make money
from it.  I have grave doubts about the ability to hold the meaning of
a domain name label stable, and I've yet to see evidence presented
that such meaning is a sound basis for a programme of treating
different strings as the same.

> I am not sure
> if semantically similar strings are actually variant strings the way we
> define variant strings, but I agree with Vladimir that on a conceptual level
> this is a case of similarity and will need to be addressed. 

And as I suggested in my mail, it is indeed an important issue and
needs to be considered.  I am suggesting, however, that it needs to be
treated separately; I was also attempting to point out the difficulty
of dealing with this case.

Finally, I was attempting to point out that there is a fundamental
dissimilarity in the traditions here for the DNS.  In the case of
different Abstract Characters that are encoded with more than one
Valid Code Point, the plain fact is that different input methods
appear to result in the same string; but as a matter of fact, those
strings are not equivalent.  This situation does not have a precedent
under the traditional LDH rules.  The case of two strings that have
the same meaning when treated as words in some language, however,
_does_ have a precedent: we treat them as different labels.  I want to
know why that tradition should be thrown over when we introduce some
new code points.

Best regards,

Andrew

-- 
Andrew Sullivan
ajs at anvilwalrusden.com