ICANN/GNSO GNSO Email List Archives

[ga]


<<< Chronological Index >>>    <<< Thread Index >>>

RE: [ga] More on localized tlds

  • To: "Karl Auerbach" <karl@xxxxxxxxxxxx>, "Danny Younger" <dannyyounger@xxxxxxxxx>
  • Subject: RE: [ga] More on localized tlds
  • From: "Paul Stahura" <stahura@xxxxxxxx>
  • Date: Mon, 26 Dec 2005 19:22:50 -0800
  • Cc: <ga@xxxxxxxxxxxxxx>
  • Sender: owner-ga@xxxxxxxxxxxxxx
  • Thread-index: AcYKj6xHVQ3yhMDFT5u6lcLMC5h9CgAAKjHw
  • Thread-topic: [ga] More on localized tlds

Karl, I think you are correct ("out of date") regarding digits in SLDs 
but not in TLDs.  

I'm not sure where the very latest rfc is either, but
if TLDs can be comprised of all digits, such as ".212", then wouldn't
that domain name, for example "32.12.32.212" would be confused with IP
addresses by all kinds of software such as browsers, ping, nslookup,
Sendmail etc? These programs would not attempt to use DNS to resolve the
name "32.12.32.212" to an IP address; they would just assume it WAS an
IP address and not do any DNS lookup.  
I guess I'm confused as to how these programs would NOT be confused. 

The requirement that the first character in a TLD string be a letter
means the TLD string cannot be comprised solely of digits.  I suppose
there could be a different requirement that accomplishes the same thing
(such as the TLD string cannot end in a digit, or it if it is
all-digits, it must be greater than 255, or if the TLD is comprised
solely of digits, then the SLD must contain a letter, or something, but
still I bet that will break some software).  




-----Original Message-----
From: Karl Auerbach [mailto:karl@xxxxxxxxxxxx] 
Sent: Monday, December 26, 2005 6:46 PM
To: Danny Younger
Cc: ga@xxxxxxxxxxxxxx; Paul Stahura
Subject: Re: [ga] More on localized tlds


On Mon, 26 Dec 2005, Danny Younger wrote:

> On the topic of localized TLDs, Paul Stahura wrote to
> me with the following observation after reading my
> post on area code tlds (reprinted with permission):

> numerical TLDs are not allowed (its in rfc 1035...

> "2.3.1. Preferred name syntax

> The DNS specifications attempt to be as general as possible in the
rules 
> for constructing domain names.

> ....

> They must start with a letter, end with a letter or digit, and have as

> interior characters only letters, digits, and hyphen."

That requirement to begin only with a letter has been out of date ever 
since the formation of 3com (1980?) and other companies (2wire) that
begin 
their name with a digit.  I'm not sure where the more up-to-date stuff
is, 
possibly in the host requirements rfc?

The DNS protocol itself can be 8-bit clean, meaning that it can carry 
Unicode.  But it has never been entirely clear to me whether 
implementations are, in fact, 8-bit clean.

If one wants to experiment with names with binary-data do a dig on:
 	maps-to-nonascii.cavebear.com
(This blows some [e.g. Fedora Core] implementations of gethostbyname())

Of course for more fun you can read the Magna Carta via DNS:
 	dig @lear.cavebear.com mc.cavebear.com axfr

 		--karl--










<<< Chronological Index >>>    <<< Thread Index >>>