[ih] Cisco origins (Was: when did APRANET -TIPs become known as -TACs)
Jack Haverty
jack at 3kitty.org
Tue Sep 30 15:29:00 PDT 2025
It's good to see a discussion like this. The list has traditionally
been focussed on the evolution of technology, algorithms, protocols, and
such. But the non-technical parts of history may actually have as
great, or greater, role in what happened over the years I never had
heard the cisco/Stanford story for example until now.
I can add a bit about the BBN router history. Note that this is my
personal recollections, others may have different perspectives.
Several of us at BBN had migrated to BBN from Licklider's group at MIT.
Lick was a strong proponent of networking, and especially his vision of
an "intergalactic network" in which humans would have their own machines
somehow all able to talk to each other, and would help people do
whatever they needed to do in life - all the time, even when the human
wasn't "logged in". Lick spent time also at ARPA, where he promoted
the same vision for militaryuse and was a key player in getting the
ARPANET going.
At the time, computers were big and expensive, but workstations had
started to appear. Our group received a bunch of machines from a
startup in a Boston suburb called the Imlac Corporation. We had a bunch
of Imlac PDS-1D machines, which were personal workstations - a desk with
computer in the base, a keyboard, and a console with lots of lights and
switches. The computer was a multiprocessor, pretty rare for that era,
containing both a CPU as was common and a Graphics Processor with its
own set of drawing instructions, both sharing the same memory space.
Mice weren't common yet circa early 1970s, although we did have a
"mouse" with 5-finger keyboard, aka Engelbart's "chord keyboard" (which
I never managed to master).
Our PDP-10 was always overloaded, so we figured out ways to use the
Imlacs to offload the PDP-10. Today we'd call it "client/server" with
the PDP-10 as a server and the Imlacs as clients. But I don't think
that term had been created yet. Or perhaps I had just never heard it yet.
Bob Metcalfe was in an office down the hall. He had built the 1822
interface for our PDP-10 and was finishing up his thesis on "Thin Wire
Interprocess Communications". He was on his way West to Silicon Valley,
first at PARC and later 3COM. Ethernet did not yet exist.
Still, to use the benefits of the Imlacs' power, we needed some kind of
LAN (also a term probably not yet created). An undergraduate showed up,
and brought some Imlac programs with him -- in particular Maze Wars.
Maze ran on an Imlac. But we had a bunch of Imlacs, so the obvious next
step was a multi-player game. No LANs, no Ethernet, not enough IMP
ports or any 1822 Imlac interfaces, but we had RS232 lines, So we
created a "LAN" using the RS232 connections and the server code on the
PDP-10. I recall figuring out ways to boost the speed of RS-232 as
high as I could, well beyond the specs, achieving almost 100 kb/second.
Low latency was crucial for gameplay.
At BBN circa 1980, we expats from MIT remembered the experience at MIT
and Lick's vision. LANs had started to proliferate, but competition
raged for the best approach. Ethernet had not yet won, and Token Ring
was a strong competitor. Still it was clear that workstations (and
eventually PCs) as well as LANs were coming, and IMP ports weren't a
viable way to connect all those personal machines to a network. Lick's
vision was congealing.
So, we tried to convince management that adding LAN ports to IMPs was a
good idea. The proposal was called "IMP Local Nets". BBN management's
reaction was basically "Great idea! Get the government to pay for it."
The government similarly said "Great idea! You guys should just do that
as a product and we'll buy a lot of them." This was sometime in the
very early 1980s IIRC.
At the time, the gateways we were operating were based on PDP-11 or
LSI-11 hardware. So a first step to offering "IMP Local Net" capability
could have been just putting an LSI-11 into some free space in the IMP
cabinet. But that never happened.
A decade later, BBN had finally created a router product, based on its
homegrown Butterfly hardware. It was shown at Interop somewhere around
1989. But the cisco Systems router was smaller, much less expensive
(IIRC about 1/5th of the price), and pretty much provided the same
capablility.
Ever since, I've been wondering if part of the historical story involves
cultural differences. East Coast and West Coast (specifically Silicon
Valley) operated very differently. Maybe that had something to do with
the influence of the Venture Capitalist crowd on Sand Hill Road.
Whatever the case, it seems that a lot of Internet-era startups came out
of Silicon Valley - Sun, 3COM, etc. In contrast, the East Coast
networking companies have mostly vanished rather quickly. Imlac,
Proteon, Wellfleet, BBN -- not names you heard as the Internet
exploded. PARC was a Silicon Valley operation with good technology, but
perhaps the East Coast corporate headquarters was an obstacle.
Earlier I described the meeting I attended at DDN to help them as a
consultant in evaluating a proposal for adding LANs to DDN. I think
the presenters were probably Len and Sandy from cisco. I never got
called to any other such meeting, although there were other product
offering they might have used - e.g., from Proteon, or Wellfleet.
Perhaps those proposals were also considered. Or perhaps the East Coast
culture wasn't as aggressive as Silicon Valley in creating startups and
promoting their products?
Maybe someone else knows that part of the history....
Jack Haverty
On 9/30/25 13:30, Guy Almes via Internet-history wrote:
> Hi Barbara,
> My memory of conversations at meetings of the NSFnet regionals was
> that Wellfleet was discussed, but I don't recall any specific of those
> regionals using them.
> As for BBN, I don't recall their routers even being discussed.
> (Although, years later, BBN actually acquired several of the regionals.)
> -- Guy
>
> On 9/30/25 4:07 PM, Barbara Denny via Internet-history wrote:
>> Any comments about why the BBN router didn't seem to become a part
>> of this story? I have heard but haven't been able to confirm that
>> 3com also had a router product at this point in time. Then I think
>> there was also Wellfleet. Did you find these routers weren't suitable?
>> I remember asking for a router build but I don't remember what I had
>> planned to do with it. It may have been just for use in the lab (and
>> maybe had something to do with Ethernet interfaces if that makes
>> sense). It is cool you found the request.
>> barbara
>> On Tuesday, September 30, 2025 at 11:53:52 AM PDT, Guy Almes via
>> Internet-history <internet-history at elists.isoc.org> wrote:
>> Noel,
>> Thanks. So early 1984.
>>
>> All,
>> The timeline seems roughly as follows:
>> <> very early 1980s: Andy Bechtolsheim designs the 68000-based
>> workstation
>> <> 1983-84: challenged by Ralph Gorin, director of Stanford's computer
>> facilities, a team adapts the workstation, minus bit-mapped display but
>> plus 3Mbps Ethernet cards, to become a router referred to as the Blue
>> Box. William Yeager is generally credited as providing the software.
>> This router software, interestingly, is multi-protocol with at least
>> both IP and PUP supported.
>> <> spring 1984: Noel infects Len Bosack with routers like the Blue Box
>> becoming a serious product
>> <> December 1984: cisco is incorporated by Len Bosack and Sandy Lerner
>> <> 1985: Stanford formalizes the "Stanford University Network". With a
>> mandate for the network to be IP-only, Len Bosack and Kirk Lougheed take
>> over the software from Yeager.
>> <> 1985: Len asks Stanford for permission to commercialize the Blue
>> Boxes. Stanford says 'no'.
>> <> So by late 1985, Blue Boxes are proliferating on the Stanford campus
>> and cisco is building a few despite Stanford's 'no'.
>> <> 1986: more Blue Boxes on campus and also cisco productizing them. As
>> Stanford staff, Len and Kirk are improving the Blue Box software by day
>> and, as cisco, are continuing to improve that same software by night.
>> Les Earnest, Len's boss at Stanford Computer Science, finds out about it
>> and confronts Len. There may be multiple versions of what happens next,
>> but before long Len and Kirk only work for cisco.
>> <> spring 1987: cisco and Stanford come to an amicable arrangement.
>> This arrangement removed what would otherwise have been an obstacle to
>> universities and others buying the resulting cisco AGS router.
>>
>> This timing is interesting to me, partly because, it meant that cisco
>> routers were available, with pretty good maturity and with no legal
>> cloud, in time for several NSFnet-related regional networks, such the
>> Sesquinet effort at Rice University where I was at the time.
>>
>> The NSFnet networks (not the backbone, but the regionals and the
>> campuses) were able to grow explosively, in part, due to the
>> availability of two good routers, the majority from Proteon and a
>> smaller number (initially) from cisco.
>>
>> In this cast of characters, several contributed to the Internet in a
>> variety of ways, but the little drama and solid innovation at Stanford
>> played a very key role.
>> None of this is new.
>> But for reasons that relate largely to the legal dispute, tellings of
>> the story often suppressed to avoid embarrassing a major corporation.
>>
>> I'd be interested in any corrections or improvements to the outline
>> above.
>> -- Guy
>>
>> On 9/30/25 12:59 PM, Noel Chiappa wrote:
>>> > From: Guy Almes
>>>
>>> > Also, do you remember when the visit to Stanford that you
>>> described
>>> > here took place?
>>>
>>> No, but there are some original contemporary clues that _might_ let
>>> us take a
>>> guess at about when it might have been. I say 'might' because I was
>>> in and
>>> out of Stanford back then a fair amount, tweaking on 'Golden', along
>>> with
>>> some other stuff going on (below) - and I have no idea _which_
>>> Sanford visit
>>> it occurred on, although I can rule one out. (Readers who aren't
>>> interested
>>> in the gory details should skip to the bottom paragraphs.)
>>>
>>>
>>> First, I got caught in INS paperwork in (I think) late 1982, and
>>> spent a good
>>> chunk of time out of the US (first in Antigua, and later in Bermuda)
>>> waiting
>>> for my permanent residency paperwork to come through. My vague memory
>>> (supplemented by the log below - my initial memory had some errors)
>>> is that
>>> the Stanford-MIT connection was starting while I was gone, and I
>>> started to
>>> work on it again after I came back.
>>>
>>> I also did some consulting for Bridge (later bought by 3Com), just
>>> after I
>>> got back, and my meeting with Len may have happened during that. (I
>>> distinctly recall that to test the Bridge work, since they had no
>>> Internet
>>> connection - indeed, no TCP/IP _anything_ - we physically dragged a
>>> Bridge
>>> box up to Stanford, and ran a whole bunch of traffic through it,
>>> after which
>>> the Bridge people were able to verify that my code had no memory
>>> leaks. It
>>> definitely wasn't that visit, though.) I did also go to Stanford on
>>> my own
>>> for some testing, prior to that.
>>>
>>> I'm too lazy to go find the Bridge paperwork (which I do still
>>> have), but an
>>> old passport reveals that I got my permanent residency in February,
>>> 1984. (I
>>> was back in the US for a short priod in the summer of 1983, too.)
>>>
>>>
>>> Second, looking through the filesystem of the MIT-CSR machine (which
>>> I was
>>> clever enough to save), I find a file with the following fragmentary
>>> notes
>>> about Golden (I also see that in March, 1984 Liza Martin built a C
>>> Gateway
>>> for Barbara :-):
>>>
>>> /* History of changes to Golden Gate config. */
>>>
>>> Who When What
>>> Bob 8/17 Added fix for Apranet deafness bug, made load
>>> module start
>>> in CGW rather than DDT, made ISI the default gateway rather
>>> than BBN.
>>> Bob 8/19 Rebuilt to use inga.h.
>>> jnc 9/14 Removed ETHWCH (claimed not needed), removed
>>> logger (not
>>> working), added MOSDDT, made MILSRI the default gateway,
>>> made MONTR output to console automatically on startup,
>>> redid inga organization, put in version II KPALV (not
>>> yet tested, though), minor cosmetic changes to some
>>> error messages, real fix to ARPANet DMA board output
>>> hangup
>>> jnc 9/16 Adjusted static buffer allocations, made larger
>>> buffers,
>>> changed build files to be able to build versions with
>>> and without DDT
>>> Bob 10/25 Modified infwd.c to drop packets that try to
>>> broadcast
>>> out the same interface they arrived on.
>>> Modified amakgat files to include local infwd.o, this should
>>> be removed after testing.
>>> Bob 11/29 Added Interland interface. Installed infwd.c,
>>> removed
>>> makgat files.
>>> Bob 1/28/84 Included Mogul's moseth-1 to fix back-to-back
>>> packets using
>>> hardware timer. Also installed a new inga.t and su-inga.h.
>>> jnc 2/3/84 Diked MOSDDT and symbol table tro give more buffers.
>>> Use the 'nost' option to makgg, which calls samakgat.
>>> jnc 14/4/84 Fixed to load with IN cataclysym.
>>> jnc 25/4/84 Added default subnet gateway in IN.
>>>
>>> Another file (a copy of the 3Mbit Ethernet interface driver, which Jeff
>>> Mogul, before this from MIT, at this time of Stanford) reveals that
>>> the first
>>> few entries here, which have no year, are 1983. Note the "Added
>>> Interlan[]
>>> interface"; looking at the early configuration files, Golden
>>> originally had
>>> only a 3Mbit Ethernet interface (the Xerox board). So the Interlan
>>> 10Mbit
>>> board (the first 10Mbit Ethernet board available for the PDP-11) was
>>> only
>>> added in November, 1983.
>>>
>>> Note also the "Diked MOSDDT and symbol table t[]o give more
>>> buffers." Early
>>> versions of the C Gateway kept everything in the low 56KB; I later
>>> modified
>>> the code to use the PDP-11 memory mapping, in a primitive way, to
>>> keep the
>>> buffers themselves in high memory - after which there was plenty of low
>>> memory -> no need to skimp on it. I _think_ Stanford got that - but
>>> I am not
>>> certain. (I'm pretty sure that work was all done on the Proteon
>>> time-sharing
>>> machine, which is long gone, alas.)
>>>
>>>
>>> So, which Stanford visit was it? I'm quite sure I had my 'mucho $$$ in
>>> routers' brainwave on the beach in Antigua - i.e. early in 1983. I
>>> think I
>>> told Len about it fairly early - so probably in early 1984, when I
>>> was at
>>> Stanford, as part of the Bridge project. But don't rely too hard on
>>> that.
>>>
>>> It would be interesting if someone asked Len if he remembers that.
>>> It's quite
>>> possible that he genuinely doesn't; memory is odd, that way. _I_
>>> remember it,
>>> I think, because I have often thought of it - especially soon after it
>>> happened, because Cisco was such tough competition for Proteon.
>>>
>>> Noel
>>
>>
>> --
>> Internet-history mailing list
>> Internet-history at elists.isoc.org
>> https://urldefense.com/v3/__https://elists.isoc.org/mailman/listinfo/
>> internet-history__;!!KwNVnqRv!FJFwSyWB-
>> Yun2_oBq3OEOMkJ13dbfvHbObRllvaEsJuPvXlseKdyBdQeeuBEQ72WHQ18mIWtXOnP9vnnL1buZEqP8lAZgw$
>> <https://urldefense.com/v3/__https://elists.isoc.org/mailman/listinfo/internet-history__;!!KwNVnqRv!FJFwSyWB-Yun2_oBq3OEOMkJ13dbfvHbObRllvaEsJuPvXlseKdyBdQeeuBEQ72WHQ18mIWtXOnP9vnnL1buZEqP8lAZgw$>
>>
>> -
>> Unsubscribe:https://urldefense.com/v3/__https://app.smartsheet.com/b/
>> form/9b6ef0621638436ab0a9b23cb0668b0b?
>> The*20list*20to*20be*20unsubscribed*20from=Internet-history__;JSUlJSU!!
>> KwNVnqRv!FJFwSyWB-
>> Yun2_oBq3OEOMkJ13dbfvHbObRllvaEsJuPvXlseKdyBdQeeuBEQ72WHQ18mIWtXOnP9vnnL1buZEpbi_fvEA$
>> <https://urldefense.com/v3/__https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The*20list*20to*20be*20unsubscribed*20from=Internet-history__;JSUlJSU!!KwNVnqRv!FJFwSyWB-Yun2_oBq3OEOMkJ13dbfvHbObRllvaEsJuPvXlseKdyBdQeeuBEQ72WHQ18mIWtXOnP9vnnL1buZEpbi_fvEA$>
>>
>>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20250930/93bc2cf6/attachment-0001.asc>
More information about the Internet-history
mailing list