[TUHS] Of login (and host) names

Norman Wilson norman at oclsc.org
Tue Jul 19 00:35:14 AEST 2016


Google was not the first place Rob and Dave had fun with names.

At one point, Rob had a duplicate entry in /etc/passwd,
with login name r, password empty, normal userid/groupid/home
directory, special shell.  The shell program checked whether
it was running on a particular host and a particular hardwired
serial line: if yes, it ran the program that started the Research
version of the window system for our bitmapped terminals;
otherwise it just exited.  The idea seemed to be to let him
log in quickly in his office.

I think that by the time I arrived at Bell Labs he'd stopped
using it, because it no longer worked, because we no longer
ran serial lines directly from computers to offices--everyone
was connected via serial-port Datakit instead.

While I was there, senior management bought a Cray X-MP/24 for
the research group.  (Thank you for using AT&T.)  Since it too
was accessible via Datakit (using a custom hardware interface
built by Alan Kaplan, but that's another story), it had to have
a hostname.  It was either Dave or Rob, I forget which, who
suggested 3k, because (a) it was a supercomputer, so `big bang'
seemed to fit; (b) it was Arno Penzias, then VP for Research,
who got us the money, so `big bang' and 3K radiation seemed
even more appropriate; and, most important, (c) it was fun to
see whether a hostname beginning with a digit broke anything.

So far as I recall, nothing broke.  Some people who were
involved with TCP/IP networking at the labs were frightened
about it; I don't remember whether that Cray was ever connected
to an IP network so I don't know whether anything went wrong
there.  Of course such names are not a problem today, but
in those long-lost days when nobody worried much about buffer
overflows either, such bugs were much more common.  Weren't they?

Norman Wilson
Toronto ON


More information about the TUHS mailing list