[ih] Domain Names

[Noel Chiappa]

A couple of points...

    > From: Craig Partridge <craig at aland.bbn.com>

    > There was a text table for hostnames ... It was managed by the SRI NIC.

Once non-ARPA networking stuff started appearing, there were also some sites
which had 'auxilliary' local host name tables, which were (at MIT at least)
automatically merged with the NIC table and the resulting 'augmented' table
was propogated locally.

The extra hosts were usually workstations (i.e. clients only, not servers for
anything), so the fact that they weren't into the NIC table was not a problem;
nobody would be connecting to them anyway - unlike in the early days of the
network, when almost all hosts were servers as well as clients. The
client/server model is of course now ubiquitous...

    > Note the fact that hosts periodically got deleted by partial transfers,
    > software bugs, etc. is why many Internet protocols permit use of raw IP
    > addresses. It also explains why many old Internet folks still remember
    > the IP addresses of some machines 

A big reason for the numeric input for IP addresses, for us, was that we had a
number of test machines of various kinds which didn't appear in the host-name
tables (too much hassle to add them), and you needed to be able to specify
those test machines.

And in the early stages of writing network software, usually the code to deal
with symbolic host-names was usually one of the latter things that got
written, you started out with only the numeric interface, so that generally
got left in when you later added host-names.


    > periodically put together Internet maps, I think mostly to help NOC
    > folks as they interacted with the rest of the Net. 

Do note that these maps (perhaps because they were seemingly intended as they
were for the people running the 'core' routers) didn't really show the
_whole_ Internet (i.e. all the IP routers, and separate networks), but only
what BBN could easily see and/or was important to BBN.

Things that were missing included all the back-door connections, all
the internal topology (at places like MIT, which showed up on that map
as a single node), etc.


    > From: Tony Li <tony.li at tony.li>

    > Certain operating systems also loaded the hosts file into memory

There was also a UDP-based 'name lookup' service (which some machines
at MIT actually used) which allowed machines without disk storage to
support host-names in the user interface.

	Noel