Re: [dnsext] we need help to make names the same, was draft-yao-dnsext-identical-resolution-02 comment

[Eric Brunner-Williams <[e-mail]>]

On 2/22/11 8:40 AM, Vaggelis Segredakis wrote:
> Hi Eric,
>
>> err, well, yes in principle, and no in practice. that is, a series of
>> pick-according-to-taste-or-belief{blunders, errors, accidents, acts of
>> genius} has lead to some specific use cases. han sc/tc. greek tonos.
>> failing to solve leads to unintended consequences.
>
> I couldn't agree more with your last sentence above. Before landing in this
> WG with this issue I tried to explain the situation back when the first
> protocol of IDNs was designed. The belief then that a protocol should not
> take into consideration specific issues but rather stick to the "domains are
> not words" motto brought us all ten years later to the point of discussing
> adding rr types in DNS, rather than designing correctly a translating
> application-layer protocol right from the beginning, with formatting data
> etc to solve all problems.

It is peculiar that we have both the uninterpreted 8 bits semantics, 
however unfashionable binary labels are presently, and the xn-- prefix 
to signal an encoding within a larger repertoire of characters, and 
the per-character associated property of directionality, and probably 
some other stuff that escapes me writing this will on a concall, yet 
we have nothing (I can think of) that provides a description of 
(sub)string (sequence of character) semantics (other than the charming 
property of "." as construed by the UTC, of directionality, which it 
happily associates with proximal characters, effecting directional 
leakage across label boundaries in bidi scripts).

A source of requirements has been primarily concerned with the visual 
similarity of some characters in larger repertoire in a single name 
space. While this appears to be a "sub-word" concern, as the problem 
to be solved is significantly less common absent outside of that 
single name space, it is confusing that this is not viewed as an 
instance of the "domain is a word" frame of reference.

I mention this to point out that the sources of requirements have 
addressed issues other than those identified in this thread.

Eric
_______________________________________________
dnsext mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/dnsext