Re: Combining characters (was: Re: [idn] hostname historyhell)

[liana Ye <liana.ydisg@juno.com>]

Hi, Vint:

  Thanks for the encouragement, and I'd like to 
propose a more specific layering of IDN symbols:

From the top where the user input buffer offers:

 Layer 3:  label seperators and label order normalizing

 Layer 2: Bidi label normalizing (or verticle label normalizing)

 Layer 1.5: diacritic marks and combining symbol normalizing

 Layer 1:  IDN identifier matching or whatever comes out 
               of [nameprep].

The reason for Layer 1.5 is that these symbols can be 
treated in a similar way with Han characters depending on 
what architecture we end up with, and what ACE will be 
our focus. 

Liana

  
On Thu, 22 Nov 2001 13:13:56 -0800 vint cerf <vinton.g.cerf@wcom.com>
writes:
> Eric, et al,
> 
> I have been observing the vigorous discussion and
> appreciate the hard work going on here.
> 
> A key element is the purpose for domain names. Are they
> just host names or are they labels into information space?
> Of course their origin is the former but they seem to have
> morphed into something rather more elaborate.
> 
> The discussion you are having is very important both for
> the future of Internet applications and generally for the
> stability and consistency of the DNS itself.
> 
> Vint
> 
> At 10:33 AM 11/22/2001 -0600, Eric A. Hall wrote:
> 
> >"Eric A. Hall" wrote:
> >
> >> The specific concern I had is already addressed in nameprep so 
> there
> >> is no need for the exception.
> >
> >What I am concerned with is the use of domain name which consist
> entirely
> >of punctuation, or in this case, which consist entirely of 
> combining
> >marks. These represent a rules problem with the principles behind 
> the
> >"safe-set". EG, we forbid domain names like ````````.com using 
> U+0060,
> but
> >````````.com may be legal if the combining character U+0300 is used
> >instead.
> >
> >I get lost in the shrubbery of specs, so can somebody tell me 
> whether
> or
> >not this is a valid concern? I can't tell for certain what happens 
> at
> the
> >end of nameprep with such a label.
> >
> >If this is a problem, a solution might be to change the prohibition
> >against leading hyphen to a prohibition against leading hyphen and
> >combining characters.
> >
> >--
> >Eric A. Hall
> http://www.ehsco.com/
> >Internet Core Protocols
> http://www.oreilly.com/catalog/coreprot/
> 
> 
> 
>