Fwd: The Root has got an A record

Bill Stewart nonobvious at gmail.com
Thu Oct 13 02:14:03 UTC 2005


Back in the mid-80s, when some people at Bell Labs were trying to get
the rest of us there onto the DNS bandwagon, there were some people
who didn't like it.  Pike and Weinberger put out deep theoretical
papers  like The Hideous Name on relative vs. absolute names and the
effects of syntax (available at
http://cm.bell-labs.com/cm/cs/doc/85/1-05.ps.gz ), and the Plan 9
naming structure, and Honeyman and Bellovin wrote pathalias to
optimize communication paths across bang-space and other namespaces. 
I mainly grumbled about the unlikelihood of everybody being willing to
let some central authority decide whose machines could be named
gandalf and mozart given the current anarchic structure of uucp
naming, a prediction which proved resoundingly wrong over the next few
years as DNS took off like wildfire because it was obviously much more
convenient. :-)

The main feature of a global hierarchical namespace root is that
"There Can Be Only One" (Highlander, 1986).  That doesn't mean that
other people can't use the same syntax and software to describe a
different namespace that may overlap the Internet's namespace and may
resolve to the same addresses in many cases, and over the years there
have been occasional alternate-root namespaces grabbing a fraction of
a percent of the market, and sometimes they've even been administered
well enough that their few users don't all give up immediately.   But
when they do something wrong with their "root", that doesn't mean that
there's anything wrong with "the" root - it just means that their
users may get unpredictable results, which is something they're mostly
used to anyway.

The DNS namespace is designed that lots of things can be grafted under
it, and much of the DNS name resolution software is designed to
resolve local as well as global names.  So company example.com with
globally-named servers like engineering.example.com or
london.example.com can have users who refer to those servers as
"example" or "london" as long as they administer their DNS correctly. 
And Joe-Bob's Alternate Root Services can have locally-usable names
like www.example.fun which are also globally accessible as
www.example.fun.joe-bob-alt-root-example.net by people who don't use
their name resolvers (again, if they configure everything correctly) -
but many of the alternate roots over the years haven't wanted to do
that, because it makes it obvious that they're not the "real" root,
just a wannabe.

There have been other global namespaces - ICQ was very popular for a
while, and it didn't get bothered by the WIPO-and-ICANN crowd because
nobody worried too much about trademark violations in a flat numerical
namespace that didn't correlate with anything else.  On the other
hand, the ENUM proposals do have serious issues of namespace policy
and centralization-vs-decentralization - should their hierarchical
number space be forced to buy E.164 numbers from the Telco Gods? 
Should anyone who has a PBX be able to manage ENUMs for extensions
under it, and should anybody with a phone number be able to define
ENUM numbers under it (e.g. 5.4.3.2.1.0.0.0.1.5.5.5.3.2.1.1 to get
extension 12345 at +1-123-555-1000, or fax.0.0.0.1.5.5.5.3.2.1.1 to
get the fax machine?)



More information about the NANOG mailing list