Thread "[ih] A laugh and a question" at
was what inspired me to bring fuzzballs up in simh.
There were some inaccuracies in that piece as well which is why history is sometimes fun.
It was definitely December 82 Not 83 because it was a few days before LINK 0 was going to
be shut off on the APRANET and we were scrambling to convert our machines to TCP. As
Clem and I have discussed in a side conversation, it was Gurwitz’s contribution to 4.1a
rather than 4.2a (was there a 4.2a even)? I remember pouring over the 4.1a/4.1c/4.2
stuff doing the user mode conversions for our machines. I really disliked the berknet
host table format (alas this has persisted longer than ping) and there was not yet a
program that converted the table that the NIC distributed to that format. I rewrote
rhost (and it’s related functions) to read the NIC host tables directly. I also wrote a
program to periodically retrieve the table from the NIC.
Later someone provided a program to build a BSD-style host table from NIC table. We
inadvertantly blew up this program when we added a machine, coincidentally called BRL-ZAP
to the table. It was the first ethernet-connected laser printer (an Imagen Imprint-10)
we had. I just put 68000 as the CPU type without thinking about it. I got some irate
emails from people thinking I’d crashed their program intentionally given the hostname.
Apparently the conversion program used YACC and an incorrect grammer to parse the table.
They had assumed that machine types always began with a letter. To me the fact that
they had used YACC was overkill (my direct reading program didn’t do so), the file was
essentially a bunch of fields separated with COLONS (and in some fields subdivided by
commas). There was some file that every UNIX machine was already parsing continually
without the need for a grammar similarly formatted (/etc/passwd anybody?).