[TUHS] [COFF] Pondering the hosts file

George Michaelson ggm at algebras.org
Fri Mar 12 13:27:12 AEST 2021

Funny story about /etc/hosts. and HOSTS.TXT. It was SRI NIC or NIC
DDN/MIL's job to maintain and distribute. Centrex model to keep the
numbers up, and hostnames aligned. We did shell ftp fetches and local
distribution. Yellow pages was a godsend in mant ways, the operations
duty cycle to keep this stuff up to date was a bit of a PITA.

Enter yourself in HOSTS.TXT. It was sorted. Often, Alpha. So people
like BBN were up top. People like MIT were midway. UCL-CS was down
low. Oh well. Good thing we don't do linear lookup.. Or it was sorted
by IP. But, somehow no matter what sort, we were down the back. Some
things basically seemed to to linear. And, as the host count rose, be
it sorted by IP or sorted by name, hosts in UCL-CS seemed to be
"bottom" for finding. Bottom in time sense.

So: you try to telnet to a US node, (or FTP) -and it dies. BBN
butterfly has LRUd you out. You get that fixed. That takes time too.
Hopefully less often. Shame BBN was on a fixed price fixed term
contract and the EGP (pre BGP) was not patchable. Things come back

You try to telnet to a node again. You get connected. Long pause. you
get login: prompt. Long pause. You start typing login.. it drops link.

Getty/login had a 30 second timer. We were sitting on 29-31 seconds
lookup delay to resolve the client IP, to assert the login prompt.
Sucked to be us.


More information about the TUHS mailing list