Re: [idn] Web navigation for IDN resolving

["Soobok Lee" <lsb@postel.co.kr>]

----- Original Message ----- 
From: "Yves Arrouye" <yves@realnames.com>
To: "'tsenglm@????.??.tw'" <tsenglm@cc.ncu.edu.tw>; ""Yves Arrouye"" <yves_arrouye@hotmail.com>; <idn@ops.ietf.org>
Sent: Tuesday, March 26, 2002 3:40 AM
Subject: RE: [idn] Web navigation for IDN resolving


> >      As I  know in Taiwan and China Area, the input Chinese character
> > string
> > with non-protocol-head in Chinese Domain name form  typed into input  bar
> > of
> > IE , the browser will intercept these string and convert it to a UTF-8 %nn
> > string as one of parameters to auto.search.msn.com , then it will be
> > passed
> > to Realname's server. The server of Realname will convert the dotted IDN
> > domain name to an UTF-8 form and re-post it in the input bar with an error
> > reply to hold it in client.
> >      The server of Realname never to do dns lookup and not try to  re-
> > direct
> > the client browser to do dns lookup in replied UTF-8 form.
> 
> The DNS lookup happened *before* the string was URL-encoded and passed to
> Autosearch, which only happen after a failure. 

In MS IE 5.0 or later,
this is true only for these two kinds of url inputs: try dns lookup first

  1)   with  explicit http: prefix

       http://www.<ML>.com
       http://<ML>.com

  2)   without http: prefix

       www.<ML>.com

  In these two cases, IEs try dns lookup for IDN hostname regardless of whether it is in utf8 or legacy encoding.

IE does *not* try dns lookup for this third kind of url input without http: prefix

  3)   hostnames begin with non-ASCII char
       <ML>.com 
       <ML>.<ML>.com

  IN this case, those inputs are treated as invalid inputs and  passed into
       auto.search.msn.com for MSN search/Realnames keyword lookups.

This IE's heuristic may be confusing to someone, but, that is not a error
and just attempts to be loyal to both user expectations and ASCII only URI spec.

IMO, 1) and 2) dns lookup should also have been prohibited and should be passed to search engines
 for IDNA or UTF8 conversions, because in most cases, they are in legacy encodings and make 
 troubles like legacy code confusion/collisions on DNS servers without appropirate conversions.
 
Netscape has different heuristic. Netscape passes any chars(including !?#$,,,) into dns lookups. 
  
Soobok Lee


> Please refer to
> draft-arrouye-idn-ie5-resolution-00.txt. RealNames *does* provide resolution
> services for IDNs on a contractual basis, and so far we resolve .com, .org,
> .net, and .jp. IDNs typed with these TLDs will resolve using an ACE.
> 
> YA 
>