[ih] multi-protocol routers, bridges (Was: "The First Router" on Jeopardy)

[Noel Chiappa]

    > Bill Nowicki
 
    > As we soon ... cloned the MIT LSI-11 system

This will not really be of interest to people here (but maybe - see the Len
Bosack story below), but I wanted to put a corrrection in the record close to
the original.

I don't think Stanford actually "cloned the MIT LSI-11 system" (probably
because doing so would have involved getting the toolchain, which ran on the
MIT V6 PWB1 PDP-11 UNIX, running there); instead, loads for the one router at
Stanford which ran it - "Golden", Stanford's ARPANET gateway - were built at
MIT.

I recall making at least one trip out to Stanford to work on Golden
(installing metering to monitor packet queue lengths, IIRC), during which I
ran into Len Bosack, who was running the Stanford timesharing systems at that
point, and explained to him how there was going to be a big maket for
routers, and I was in the process of working with Proteon to turn out a
product. In retrospect, probably not the best idea! :-)

Jeff Mogul may also have been involved; I found his name in some log files
(below). He'd been associated with our group at LCS as an undergrad, when he
put a ring interface on the DSSR/RTS VAX; he then went to grad school at
Stanford. He would have known of the router work at MIT, and probably they
figured it was easier to use our code on their ARPANET gateway, than write
ARPANET code of their own for Bill Yeager's router.

I have a dump of that MIT V6 PWB1 PDP-11 UNIX, and just looked at the config
files. Below are some amusing/interesting snippes from a later one, after a
DIX Ethernet board had been added to it. Interestingly, although there was
PUP code available for the MIT router, Golden seems to have been IP-only
(from what I can see in the config files).

	Noel

----

  /* Node name */

  char    name[]  "Golden-Gateway";


  /* Actual network configuration table. Note the trailer word in
   * the ARPANet line; NOT a hardware problem in ANY board, but
   * IMP trailer. (See your 1822 manual, tuna.)
   * LOSERS: don't frobozz the n_max packet count fields; this
   * configuration is intended to maximize ARPANet throughput
   * in a gateway that's not computationally loaded. In particular,
   * leave the ARPANet maxip larger than the Ethernet!!!
   * "We like our defaults. If we hadn't liked them, we wouldn't
   * have made them the defaults, now, would we."
   */

  {
   { NULL, &eth_prinit, &eth_out, NULL, &eth_in, NULL, NULL,
     /* EtherNet */       2, 3, 4, 0, ETHMAX, sizeof(struct ethpkt), 0,
			  DVETH, DVETH+1, DVETH, T_ETH, C_BRD },

   { &ar_init, &ar_prinit, &ar_out, NULL, NULL, &ar_get, NULL,
     /* ARPANet */        4, 3, 0, 0, ARMAX, sizeof(arpkt), sizeof(word),
			  DVIMP, DVIMP+1, DVIMP, T_ARPA, 0 },

   { NULL, &e10_prinit, &e10_out, NULL, &e10_in, NULL, NULL,
     /* 10MB Ethernet */  1, 3, 4, 0, EMAX, sizeof(e10pkt), 0,
			  DVIE, DVIE+1, DVIE, T_E10, C_BRD },
  };


  #define NINADDR 3

  inia    iniatbl[NINADDR]

  {       { { 36, 050, 0, 076 }, NULL },          /* Ethernet */
	  { { 012, 1, 0, 11 }, NULL },            /* ARPAnet */
	  { { 36, 8, 0, 1 }, NULL },              /* 10MB Ethernet */
  };

  /* For subnet kludgery; if you are on a class A net with subnets set
   * this variable to your net number to enable subnet routing. If
   * zero, this feature is turned off.
   */

  unsb    myanet  36;