Which hosts table? The Berkeley one or the REAL internet one?
The Berkeley one (which I think may predate the IP implementation) is the one that we know as /etc/hosts that has the address then the namees of the hosts.
The "real" one is the one the NIC put out in the pre-domain days. It's defined in RFC 952, looks like
HOST : 10.0.0.29 : BRL.ARPA, BRL : PDP-11/70 : UNIX : TCP :
There was also a simple TCP service that would serve up the file.
I detested the Berkeley one and we always downloaded and used the NIC table on our machines. I rewrote "rhost" and it's successors (gethostbyname, etc...) to read directly from the NIC format.
Amusingly one day we got an Imagen ethernet-connected laser printer. Mike Muuss decided the thing should be named BRL-ZAP and since I didn't know what to put down as the machine type, and it did have a 68000 in it, I had Jake put 68000 in the entry in the host table.
The next day I got all kinds of hate mail from other BSD sites who assumed I had intentionally sabotaged the host table. Apparently, the BSD systems used a YACC grammar to parse the NIC table into the Berkeley one. The only problem is they got the grammar wrong and assumed the CPU type always began with a letter. There parse blew up on my "ZAP" host and they assumed that was the desired effect.
I countered back that using a YACC grammar for this was rediculous. There was already a real popular file on UNIX that had a bunch of fields separated by colons and commas (/etc/passwd anybody) that it was never necessary to use YACC to parse.
-Ron