[dns-operations] about the underline in hostname

[Phillip Hallam-Baker]

On Thu, May 29, 2014 at 10:29 AM, Matthew Pounsett <matt at conundrum.com> wrote:

> Host names are only allowed the characters [a-z] (case insensitive), [0-9], and [-].  See RFCs 952 and 1123.
>
> Domain names may use any string as a label, so for example the underscore is perfectly legal.  See RFC 2181.

This is the second time today I have come across the form 'very old
RFC amended by another old RFC'. And it worries me. Take the text in
RFC1123:

      The syntax of a legal Internet host name was specified in RFC-952
      [DNS:4].  One aspect of host name syntax is hereby changed: the
      restriction on the first character is relaxed to allow either a
      letter or a digit.  Host software MUST support this more liberal
      syntax.

A document from 1985 with an errata in 1989 is still the definitive
definition of host name. And that is a problem because all numeric
names are valid DNS hostnames.

So is 10.1.2.3 an IP address or is it potentially a valid DNS host name?

According to RFC952 as updated by RFC1123 it is. Even though it makes
no sense given what follows in RFC1123 which clearly expects that
there will never be all numeric TLDs and thus no possibility of a
confusion between host names and IP addresses.


There should be a standards track document that clearly prohibits all
numeric TLDs. Do we really want to have to trust that a future head of
ICANN faced with a choice between rejecting such a proposal and a
second luxury yacht decides to go sailing?

At any rate, given the state of the specs, I don't see cause for being
rude when people ask questions about them.

Take a look at the title of RFC 952:

DOD INTERNET HOST TABLE SPECIFICATION

Why would a reader expect this to be relevant to the Internet? It
looks to me as if it is a set of specifications that are local to the
US Department of Defense.


There is a document quality issue here. None of these specs would be
published as a proposed standard today.