[ih] Domain Names

[Jack Haverty]

It was pretty simple really.  Hosts (computers) in the early 70s had
names, like MIT-DMS or USC-ISIA or BBN-TENEX or UCLA-CCN or SRI-KL.
Note that there was already a bit of structure emerging - the names were
often of the form <organization>-<local name>, but that structure was
only understood by humans.  To all the computers it was just a text
string.

Subsequently the structure transformed into <localname>.<organization>,
and led to the DNS, where software could parse the name structure and
act on the contents.

In the beginning, ... there was a master file (HOSTS.TXT) that contained
all of the names and their corresponding physical addresses, which were
basically ARPANet addresses - i.e., IMP number and physical port.  The
master file was kept at SRI, which kept the most recent version
available at a well known FTP location.

People (host administrators, etc.) would download the file every so
often to get the latest addresses.  Programs like Telnet would use the
local file to change names people typed in, like "MIT-DMS", into the
appropriate numeric address for opening a network connection.  

People who brought up new hosts, or were silly enough to move them from
one physical network port to another, would have to inform the NIC of
what they had done, so the changes could be entered into the master
file.  Of course the changes wouldn't have an effect on any particular
host until that host had downloaded the latest copy of HOSTS.TXT  That
could cause lots of confusion since there was little consistency in how
frequently a particular host got a fresh copy of the hosts.txt file.

This worked fine as long as the number of hosts fit on a few pages of
printout.  But it got unwieldy as the numbers of hosts grew, and users
became less patient waiting for changes to take effect.  Plus it became
more difficult to pick a good name or "reserve" a piece of name-space so
you could, for example, have all of the NET-xxx names for your own use.

Something better had to be created.  Out of all that pressure, DNS came
into being.   You can still see vestiges of the original scheme - for
example in the /etc/hosts files on Unix systems, which look a lot like
the old hosts.txt file.  The DNS basically took the old human-driven
system and made a system that provided similar but much richer services,
able to handle much greater scale and speed.

I believe that the "!x!y!z" type of name/route originated from the days
of dial-up phone networks, in which a computer would periodically dial
specific other computers and exchange email.  The *user* specified the
specific route that a message would take.  So a name like X!Y!Z meant
that machine x would dial up y, exchange the mail, and disconnect.  Then
Y would dial Z, pass the message, and disconnect.  In those days when
phone calls were expensive and host administrators had specific ideas
about where their phone charges could be spent, such flexibility was
important.  In addition, since not just anyone was allowed to connect to
the Arpanet/Internet, this made it possible for users on non-network
hosts to be email-connected to users on Arpanet hosts - you just needed
to find an x!y!z path that would make the appropriate linkage.  

One might even argue that this "email gateway" approach was the first
internet...by providing email connectivity between two very different
networks.  Think of email messages as "IP packets".

HTH,
/Jack Haverty



On Tue, 2010-01-19 at 20:21 -0800, Richard Bennett wrote:
> We're coming up on the 25th anniversary of the first domain name 
> registration, that of symbolics.com on March 15, 1985. Not the first 
> domain name created, but the first one registered in the fledgling 
> domain name system. Since I'm too old to remember that era, I'm 
> wondering if anybody has any salient observations about what the 
> Internet was like before the domain name system was created. How did 
> people keep track of everything?
> 
> I seem to remember a cumbersome system of bang addresses for e-mail that 
> apparently arose out of UUCP, but wasn't there a more elegant system of 
> naming for ARPANET and the fledgling Internet before 1985? I have the 
> feeliing that there will be some events to commemorate the rise of the 
> Dot Com era, and it would be nice if some of the facts were more or less 
> in order.
> 
> Thanks,
> 
> RB
>