1996-09-14 - Re: Domain names - Alternic - Authority

Header Data

From: Bill Stewart <stewarts@ix.netcom.com>
To: Law & Policy of Computer Communications <CYBERIA-L@LISTSERV.AOL.COM>
Message Hash: 1b6d546581088e82f504f24ff3b1a833483605ab8b031905e264e628c340d279
Message ID: <199609140915.CAA28221@dfw-ix6.ix.netcom.com>
Reply To: N/A
UTC Datetime: 1996-09-14 11:34:26 UTC
Raw Date: Sat, 14 Sep 1996 19:34:26 +0800

Raw message

From: Bill Stewart <stewarts@ix.netcom.com>
Date: Sat, 14 Sep 1996 19:34:26 +0800
To: Law &amp; Policy of Computer Communications              <CYBERIA-L@LISTSERV.AOL.COM>
Subject: Re: Domain names - Alternic - Authority
Message-ID: <199609140915.CAA28221@dfw-ix6.ix.netcom.com>
MIME-Version: 1.0
Content-Type: text/plain


(The Top-Level Domain Name Wars have been breaking out in Cyberia-L,
not a usual place for technical discussion, plus popping up and down
in Cypherpunks.  Hope y'all don't mind me mixing this to both.)

Back when domain names first came in, a number of us didn't believe
that a centralized organization could pull it off, especially when
it was being done by a collaboration of the US military types
and the European Phone Companies, who were well known to be clueless.
The email world wasn't just the ARPAnet, it included all of UUCP-land,
PCs were starting to occupy their spaces, and there were various
DECish things and Berknets and other radically non-conformable 
architectures and overlapping namespaces.  UUCPland worked because a few 
dedicated people, mainly at Bell Labs and some universities, with lots 
of time and non-beancounted-non-billed funds, kept track of where 
machines lived, phone-polled sites that couldn't afford it, 
tried to keep track of which of the umpty-seven machines named "bilbo"
or "mozart" could talk to which others so you could occasionally
get mail to someone who lived at ucbvax!allegra!mozart!zorro!bilbo!gandalf
and know it was the same bilbo!gandalf you'd talked to last week
and maybe get mail to people reliably across complex paths.
        Pathalias was a cool program that took link data and Dijkstra'd it
        together into a relatively connected bunch of shortest paths.

The idea that Somebody Thought They Were In Charge was annoying to
many of us, especially Somebody Bureaucratic.  Peter Honeyman
set up a bunch of his machines as the .FUN domain for a while;
it was run by fun people.  Rob Pike and Ken Thompson later wrote a
marvelous technical rant called "The Hideous Name" about how
locally-based addressing makes far more sense than global -
it both reflects reality and lets the decisions about naming
and such be made by the people and processes that need to make them.

Well, domain names eventually became a raving success, just because
it was so much easier to say where someone was (administratively,
if not physically) than to guess how to get there hop-by-hop.
Separating administrative addresses from routing was an important
mental step.  Stuff like .BITNET and .CSNET were workable,
.UUCP had its real problems but eventually became sort-of-defined
relative to uunet, a machine set up as a research project that
had turned into a small business.

One of the things that made it work well was that any given machine
only had to know a manageable amount of stuff - you needed your own
subdomains, a path to the top of your organization, a way to recognize
that given patterns were high-level domains, and a path to the 
Root Servers - plus you needed somebody Higher Up to enter your
name and address in their administrative database so other people
could find you.  It's easy to find where to look among a few hundred
entries (a few technically-cooperative countries, a few zones for
non-countries, and a few hundred countries that didn't really have email :-)
Occasionally a country comes on board that your internal
nameservers didn't recognize, and you need to get your administrator
to add .nl or .kr or whatever.

>From that relatively short list, which you don't need any blazingly
complex database to administer, you can get anywhere.  The standard
DNS BIND tries to solve a bunch of complex _local_ problems and
work portably in many different environments, but doesn't need
to be tuned blazingly complexly to work almost anywhere.
The systems that need to go beyond that are the servers for .com
and to some extent .edu, which have to keep track of a lot more sites,
and places that have complex internal structures or are trying to
solve fancy problems like mobile users.   But www.joesgarage.com
doesn't need to do most of that.

I think polluting the top-level domain name space is a bad move.
While geographically-based names aren't a great choice for
representing non-small organizations, they're short, and if you
open up the entire top-level you just complicate everyone's life
and in a year or two you'll have as many people carving out space 
at the toplevel as you currently have in .com, raising the ego levels
without making the name space any bigger.  Now, I suppose there's a
certain nice philosophical position of being able to say
"You're .gov.fr - we're .microsoft, and we're at a higher level than you!"
or "You're .gov.uk - we're .gov.AnTir, and everybody in _our_
kingdom is there because they want to be!"


#			Thanks;  Bill
# Bill Stewart, +1-415-442-2215 stewarts@ix.netcom.com
# <A HREF="http://idiom.com/~wcs"> 	
# You can get PGP software outside the US at ftp.ox.ac.uk/pub/crypto






Thread