ICANN ICANN Email List Archives

[gtld-evaluation]


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

Correction to "Technical Requirements for all Labels (Strings)"

  • To: gtld-evaluation@xxxxxxxxx
  • Subject: Correction to "Technical Requirements for all Labels (Strings)"
  • From: Eric Brunner-Williams <ebw@xxxxxxxxxxxxxxxxxxxx>
  • Date: Fri, 05 Dec 2008 17:50:36 -0500

This comment is the written statement of informal remarks made to Kurt during the Paris meeting concerning digits and stability.

The text at page 2-7 is more restrictive than necessary.

A tld may, without possibility for a confusing a sequence of ASCII labels for an IPv4 address, by network infrastructure, be a digit, or a sequence of digits, or an octal or hexadecimal representation, if the following conditions are met:

If the first label is less than 255 (o377, 0xff) , then the next label must be non-numeric, or greater than 255 (o377, 0xff). If the first two labels are both less than 255 (o377, 0xff) , then the next label must be non-numeric, or greater than 255 (o377, 0xff). If the first three labels are all less than 255 (o377, 0xff) , then the next label must be non-numeric, or greater than 255 (o377, 0xff). If the first four labels are all less than 255 (o377, 0xff) , then the next label must be non-numeric, or greater than 255 (o377, 0xff).

So, "0" is a fine top-level domain, if the "0" registry only allows non-numeric second-level domains, or if it allows numeric second-level domains, has the appropriate recursive rule set so that no label sequence of length four or less is possible, where all of the labels in the label sequence are numeric and all have a value less than 255 (o377, 0xff).

There are applications which will interpret any number as an address, however we are only concerned with those which treat a sequence of two or more labels, that is, with a sequence of digits (decimal, octal or hexadecimal) separated by dots, and under what conditions a label sequence is commonly interpreted as an IPv4 address.

See INET(3), inet_ntoa() and associated Internet address manipulation routines, which appeared in 4.2BSD, documented in X/Open Networking Services Issue 5.2.

I'll submit the alternative wording if staff decides that numeric labels which do not pose technical stability issues shall be allowed.

Eric Brunner-Williams
(wearing my "co-author hat" for XPG/1 and XPG/4.2, aka "The Single Unix Specification (SUS)"

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

Privacy Policy | Terms of Service | Cookies Policy