[Simh] EXT :Re: PDP-10 simulation: DEUNA support help needed

[Rich Alderson]

> Date: Mon, 27 Apr 2015 16:21:28 -0400
> From: Timothe Litt <litt at ieee.org>

> On 27-Apr-15 14:56, Cory Smelosky wrote:

>> ...This is getting absurd.  Just how many stacks exist?!

> That I know of?  For IP?

> DEC had a TOPS-20 TCP/IP stack.  Many of the upper level protocols were
> first implemented on t20/tenex, including smtp, ftp, telnet,
> time/daytime,tftp, dns.

> BBN had a TENEX stack.  Not sure if DEC's started with it.

No.

> USC/ISI had a KA on the ARPANET; not sure what stack.

> I believe ITS and WAITS both had TCP stacks; they certainly had NCP (the
> TCP predecessor, not the DECnet)

I can't speak to ITS at the moment, those bits are buried deep under more
current projects, of which the most relevant is WAITS on a 1095 here at the
museum.

SU-AI[1] only had 3Mbit Experimental Ethernet, courtesy of a Xerox interface
card for the PDP-11.  This was installed into the 11/40 front end on the KL[2]
in order to talk a batch of Altos donated by Xerox at the same time.  This only
ever spoke PUP.[3]

WAITS, which began as the PDP-6 monitor and diverged from Tops-10 around the
time of the 4S72 monitor, supported NCP until TCP/IP came along, at which point
the lab made a clean break from NCP.  There was never any Ethernet-related code
underlying the NCP or IP protocol stacks.

I'm currently working on getting networking going on WAITS.  I have a console-
only system running, since late last fall.  It was only a month ago that one of
the SAIL alumni advising this project pointed out that WAITS never used the
MEIS (or any other Ethernet interface) for TCP/IP.  We're trying to get a Xerox
card for the front end, since we have Altos anyway, and to figure out how we're
going to put this on the Internet.

> Interfaces: The AN10 for the KA/KI/KL (it lived on the IO Bus, so on the
> KL used the DIA20).  The AN22 for the KS.  These provided the 1822
> interface to the IMPs.   I know some folks used SLIP on serial lines.=20
> Of course once the KL got the KLNIA, ethernet was the way to go.=20
> Cheaper, faster - and by then, the IMP was gone.

> I'm probably missing some, but then I really don't have time to ponder
> this right now.

Years before Digital brought out the NIA-20, there was a successful commercial
product, the Massbus-Ethernet Interface Subsystem (MEIS), from Cisco.  This was
a 10Mbit ("10base5", in modern parlance) version of a 3Mbit board set invented
at Stanford for the purpose of networking the various PDP-10 sites on campus
together.[4]  The Stanford monitor (ancestor of the Panda monitor of which so
many people are fond) supported TCP/IP as well as PUP over the MEIS, both 3Mbit
and 10Mbit; Cisco and Mark Crispin separately decided to remove the PUP code
from their monitors.[5]

                                                                Rich

[1] AKA "SAIL", but that was also the name of the laboratory and of the lab's
    extended implementation of Algol 60 with associative data structures built
    in, so this is an easier way to refer to the computer system.

[2] SU-AI started life as a PDP-6.  A KA-10 was added and made the master CPU
    when those were introduced; an IMP was attached to the KA to put SAIL on
    the ARPANET.  Later on, a KL-10 was added to the mix, becoming the master
    CPU in a tri-processor mixed system.  The PDP-6 was taken out of the mix in
    1984 when SAIL moved into the Computer Science department in Margaret Jacks
    Hall, and lost somehow after it was used at the 20th Anniversary LCG SIG in
    Anaheim; the KA was removed from the mix a couple of years later, and the
    KL-only system was retired around 1991.

[3] PARC Universal Packets, an internetworking protocol stack which informed
    the development of IP and the TCP and UDP stacks on top of it.  8-bit
    network numbers, 8-bit host numbers, for a 16-bit address field in the
    Ethernet packet.

[4] In 1984 when I went to work at Stanford's academic computing facility LOTS
    there were more than a dozen systems running TOPS-20 or TENEX on campus:
    LOTS had 4, CompSci had 2 (SCORE and SAIL), EE had 2 (Sierra and one I don't
    remember the name of), the Center for the Study of Language and Information
    had 1 or 2 (CSLI and another), the Medical School had 2 (SUMEX, a KL, and a
    KS whose name I do not remember), the Graduate School of Business had 2 (HOW
    and WHY), the Data Center had 1 (ConTEXT, IIRC), and the Institute for
    Mathematical Studies in the Social Sciences had a dual-processor KI (IMSSS,
    running TENEX).

[5] All of Cisco's business and engineering was done on DEC-20 systems for the
    first half dozen years or so.