Discuss mailing list

["John Kozlowski \(ShofarDomain.com\)" <John.Kozlowski AT ShofarDomain.com>]

The DNS protocol uses bytes, so 16 or 21 bit Unicode characters have to be 
encoded.  Using UTF8 would work fine, but there was a concern that you could 
fool someone to going to a rouge site by using look alike characters, such as 
a Cyrillic ‘Ϻ’ in place of an Latin ‘M’.  Punycode makes it obvious that you 
are using extended characters.  However, this is just another way of doing 
things when UTF8 would be great.  Fundamentally this issue of misdirection is 
probably overrated.

To experiment with this I obtained a name using Cyrillic characters.  Some 
browsers, such as IE, display punycode (http://xn--80aa3agvl7b6d.net), while 
others like Firefox display the UTF8 (http://матрёшка.net).  The is Russian 
for Matryoshka.

Personally I think IE got it wrong (have you heard this before?) since this 
is unreadable for us normal humans.  It might be nice if Firefox would 
somehow highlight the fact that extended characters are used.

Clearly as long as you are sending compatible information in the DNS query 
how you store or manage that in an application can be either punycode, UTF8, 
UTF16, UTF32, or EBCIDIC.  You might have issues with the last one.

John Kozlowski

-----Original Message-----

Would it be possible to use DNS without punycode? I realize this would be 
non-standard, and would break compatibility with DNS servers in the public 
Internet, but is there any protocol-specific reason you can't?

My understanding was that the engineers who came up with DNS decided 
allowable characters to be ASCII simply because they didn't expect the 
Internet to become international and huge.