[ih] Yasha Levine's Surveillance Valley: The Secret Military History of the Internet -- Some Questions

[Jack Haverty]

Hello Eric,

First, let me thank you for pointing us to the "Dream Machine" book.  I
tend to ignore "history of the network" books, since I've found that
they tend to describe a history sometimes quite different from what I
remember as one of the people who was actually there.

"Dream Machine" is an exception.  Lick was my thesis adviser, and
subsequently my boss when I was a member of the MIT research staff in
his group.  So there is a lot of overlap between my personal experiences
in 1969-1977 at MIT and the events and interactions chronicled in "Dream
Machine".

My reaction to Dream Machine was that it was right on target, consistent
with everything I remember (except maybe a few minor details).  I also
learned a lot, with the "back story" of the political history now
explaining some of the things that happened in those days at MIT.

Of course, I haven't read "Surveillance Valley", but I can offer some
insights into how it was back then, to maybe help you decide whether SV
is fantasy or reality.

First, I think it's important to understand that "MIT" was not a
monolith.  It had many pieces, and, like most institutions, they were
sometimes cooperative, usually competitive, and often unaware of what
the other pieces were doing.  That's especially true if you think of
"MIT" as not only the school per se, but also related pieces - Lincoln
Labs and Draper Labs being two major ones.

Lick's group was part of Project MAC, aka LCS (Laboratory for Computer
Science),  It occupied part of 545 Technology Square, along with the MIT
AI Lab.  LCS had many subgroups.  In addition, the building complex
housed an IBM research group (that did the DataComputer, which was
attached to the ARPANET), and even a stealth office of the CIA (really -
but that's another story), which I accidentally "outed" one day while
trying to run computer cables up to the roof through the elevator shaft.
 Oops.

Second, the ARPANET was not the first, or only, way to communicate
between computers at different organizations.  As an undergrad in 1968,
I had a part time job which involved running an APL facility for use in
some course (Metallurgy IIRC).  That was accessed by a dial-up line from
one of our computers to one at the IBM research center in upstate New
York.  Lots of that kind of informal ad-hoc "networking" was common.

IIRC, the machines "on the ARPANET" at MIT in the early years included
Multics, Dynamic Modelling (Lick's group where I hung out), AI, and ML
(MathLab).  Those all supported multiple projects (who could use the
ARPANET, with ARPA permission of course).  Those projects could have
been anywhere in MIT (or even elsewhere over dialups), using one of the
ARPANET-attached machines to do --- well, whatever they were doing.

I doubt anyone knows, or knew, what all of those pieces were and what
exactly they were doing back in the 60s/70s, who they were working for
(being funded by), or what kind of data they moved around.  LCS itself,
AFAIK, did not do any classified work.  Draper Labs did (I had a
parttime job there for a year or so, programming a PDP-8).  Probably
Lincoln too but I never worked there.

Also remember that this was the time of protests in the streets.  It
wasn't politically correct (or even safe) to be doing certain kinds of
work.  So anybody doing such stuff kept it quiet.

So, it's certainly possible that something interesting and controversial
got "transferred to MIT", but that's too vague a claim.  Without
details, it's too hard to tell if it might have happened.

I do remember, at the ICCC '72 debut of the ARPANET in DC, that one of
the most popular datasets being transferred around and sent to the
printer in the exhibit hall was ... a file containing a collection of
bawdy limericks.  Possibly the debut of "Internet Porn"?  I still have a
yellowed listing as a souvenir.

So, the lesson is that nobody could really tell what was going across
the ARPANET.....

Third, there was a behavioral pattern which I'll call "Elephant
Syndrome" -- after that old story about a group of blind people
describing an elephant based on which piece of the elephant's anatomy
each of them can touch.

Here's an example of Elephant Syndrome (ES) from Lick's group.

At one point in the mid-70s Lick reluctantly disappeared for a while to
go back to ARPA for a year.  This is described in "Dream Machine" - and
after reading that I now know what was really going on!

Subsequently, we worker bees were cajoled into being excited about a new
project - how to get our poor overworked PDP-10 to not only decode, but
to understand hand sent Morse code, and even hopefully be able to
participate in conversations with other human-manned stations.  Today,
we would call this an "expert system", but I don't think that term had
been invented yet.  I was a ham operator in high school, and got pretty
good with Morse.  So I was "the expert", and our group tried to make the
computer do things the way I did them as a Morse operator.  I would
explain how I did things as a Morse operator, and then we'd figure out a
way to get the computer to do the same.

Having now read "Dream Machine", I realize that this was driven by
Heilmeyer's (and no doubt others) desire to make the funded research
more immediately relevant to military needs.  Foreign forces were still
communicating using hand-sent Morse.   Inquiring minds wanted to know
what they were talking about.  Soon.

I remember one day when "the government" came to visit.  Half-a-dozen
men in dark suits, very serious.  I never knew who they were.  We showed
them what the system we had built could do and they seemed very
impressed.  We also advised them that we couldn't quite get "realtime"
understanding of hand-sent Morse.  The PDP-10 (IIRC - 512MB of memory,
.001 GHz CPU)! simply didn't have the horsepower.  So having the
computer interact with a live human in a Morse conversation was not
possible with our equipment.  They didn't seem concerned about that.
I think we missed an opportunity then - to ask for a few more PDP-10
systems (only a few million each..) to continue the work.  I bet we
would have gotten them with no fuss at all.   Where to put them though
-- that would have been a real problem.

That was of course the "military/surveillance" view of the Elephant and
why it was being built.

>From the MIT researchers' perspective, the Morse project was interesting
because Morse is a very very simple language.  There are only two
syllables - "dot" and "dash".  Much simpler than spoken English or any
other voiced language, but still rich in details to be handled -
dialects, accents, noisy and "cocktail-party" environments, etc.  But
building a system which could truly understand that simple spoken
language seemed like a good first step in research toward eventually
getting a computer to understand spoken conversational human voice -
something which seems now pretty close to solved, 43 years later.  Maybe.

So the "research/academia" view of that Elephant was quite different
from the military.  Same Elephant, two ideas of what it was all about.

I never saw whatever was the written proposal from MIT to somewhere in
the government to start that More "Natural Language Research/Morse
Surveillance" project.  It would be interesting to see what perspective
of that Elephant was portrayed in the proposal.  The Morse system could
certainly have been accessed over the ARPANET although I don't recall
anyone ever actually doing that.  One could certainly have promoted the
project as "aiding the surveillance mission".

Another project at MIT in Lick's group in the 70s focused on "Electronic
Messaging", which fit right in with Lick's "Dream" and we built a system
that mimicked a typical office environment.  Then we adapted the system
as part of the "Military Messaging Experiment", which was a testbed
deployed (at CINCPAC IIRC), to show how Electronic Messaging could be
used to improve the machinery of military communications - everything
from logistics to command and control in a multi-level secure environment.

Again, an Elephant which looks quite different to different audiences.

Throughout those years at MIT, I never heard of "The Cambridge Project"
or "Project CAM".  But perhaps I was looking at a different part of the
Elephant.

Is it possible that the "Cambridge Project" was something involving the
DataComputer?  It was across the hall from Lick's world at MIT, and was
on the ARPANET in the early/mid 70s.  I did several projects trying to
use the DataComputer, e.g., as an archival trusted repository for
important email (a sort of escrow agent).  I imagine there may have been
some kind of joint project between IBM and some part of MIT to use the
DataComputer in some way.   After all, IBM probably put it in a building
on the MIT Campus for some reason....

In 1977, I left MIT and joined BBN in the same group that built and was
running the ARPANET.  My first assignment was to implement TCP for Unix.
 For the next 13 years, with titles ranging from Computer Scientist to
Chief Network Architect, most of the work I did at BBN involved the
Internet in one way or another.

If you've read the "Dream Machine", you'll understand the synergy
between Lick's MIT activity, BBN, and ARPA.  The same culture and
behavior could be found in all.

In particular, we got pretty good at creating Elephants.

One example I can remember clearly.  The Internet was getting a bit
unwieldy, and the basic technology really didn't have any mechanisms for
management - things like fault detection, isolation, configuration
control, and the like.

So, a project was put together, drawing funding from several very
different sources.  To ARPA, we were working on forward-thinking ideas
for "Automated Network Management", which fit in with their mission of
advanced, risky research.  To DCA, we were working on new operations
tools and procedures which were critical to getting the users onto the
DDN, and making the Internet reliable in military environments.

We even had funding from some of the "user" communities, who had begun
building their own private clones of "The Internet" and were frantically
searching for the management tools.

Each of those groups had a different perspective on the Elephant they
were funding.  Same Elephant.

Those "user communities" even extended into the non-government arena.
By the late 80s, corporations with an early-adopter gene had gotten
tired of waiting for the PTTs and Big Iron vendors to deliver on their
promises, and were tentatively pushing forward with the TCP route,
simply by building their own personal private corporate Internet.

I consulted for one firm - a major Wall Street investment house - with a
private Internet (IIRC we called them Intranets then), with multiple T-1
lines interconnecting New York, London, and Tokyo as their own neonatal
private Internet.

They were so serious about reliability and zero-downtime that they even
had two massive datacenters, fully redundant and linked so that one
could take over if the other failed.  When I asked what situation they
were worried about, the CIO pointed out that one datacenter was on the
flight path for Newark airport.  The risk of an unfortunate encounter
with an airliner was just too high.  Each minute of downtime meant
millions of dollars lost.  A clone datacenter was cheap insurance.

When you think about it, the government and especially military focus on
C3I - Communications, Command, Control, and Intelligence.  Looking at
corporations like that Wall Street world, they do exactly the same
things - gather data, turn it into information, get it to decision
makers, and issue commands, and control their agents activities around
the world.

When the military C3I system fails, people can die.  When the corporate
system fails, billions can be lost.

The Internet is an Elephant that serves all those needs.  What it is
depends on your perspective....

So, was the Internet "from the beginning, grown from "counterinsurgency"
and surveillance operations"?

Probably there are people who think so. They may be right. You decide...

I think we were building The Ultimate Elephant.  It is what it looks
like ... to you.

/Jack Haverty






On 04/13/2018 02:47 PM, Eric Gade wrote:
> Hello list members,
> 
> Please excuse the length of this email.
> 
> I am in the process of writing a review of Yasha Levine's new history of
> the Internet, "Surveillance Valley: The Secret Military History of the
> Internet." His overall thesis is that the development of the Internet
> has, from the beginning, grown from "counterinsurgency" and surveillance
> operations, and that these aspects have not been adequately chronicled
> in other histories. Many of his claims about the early ARPA work I have
> not encountered before, and I imagine that some would find them provocative.
> 
> I'm hoping there are members of this list with knowledge about these
> claims who can help me clarify a few points:
> 
>  1. Levine asserts that there was some overlap or relationship between
>     William Godel's Project Agile and work conducted by the ARPA Command
>     and Control division under Licklider. He pulls a lot from Sharon
>     Weinberger's recent book ("The Imagineers of War") in discussing
>     both Godel and the potential connection. He writes, "[Licklider's]
>     work at ARPA was part of the military's larger counterinsurgency
>     efforts and directly overlapped with William Godel's Project Agile."
>     (52). In making this statement he actually cites Weinberger's
>     prologue, in which she says "Godel personally signed off on the
>     first computer-networking study, giving it money from his Vietnam
>     budget." It appears Weinberger is herself citing this document:
>     (https://archivesdeclassification.files.wordpress.com/2011/07/arpa-order-internet.pdf).
>     It is a part of a series that may still be classified (I have the
>     NDC looking into it). My question about this is: was there really
>     any kind of working relationship? What does this transfer of funds
>     represent? And perhaps more broadly: to what extent was ARPA
>     C&C/IPTO involved in counterinsurgency data collection and processing?
>  2. A large section of the early history in this book deals with the
>     Cambridge Project (aka Project CAM) at MIT and controversy
>     surrounding it at the time. I am awaiting a copy of the original
>     proposal from MIT (it might not come in before deadline; should
>     anyone on this list have a copy I'd really appreciate it). Levine
>     asserts that the project "would directly aid the agency's
>     counterinsurgency mission." He claims that the work of the project
>     "could be accessed from any computer with an ARPANET connection"
>     (68) and that "It was a kind of stripped down 1960s version of
>     Palantir, the powerful data mining, surveillance, and prediction
>     software the military and intelligence planners use today." He goes
>     on: "the project was customized to the military's needs, with
>     particular focus on fighting insurgencies and rolling back communism
>     [...] It was clear that the Cambridge Project wasn't just a tool for
>     research, it was a counterinsurgency technology." (68-69)
> 
>     Is that not an accurate description of the proposal? Were any
>     members on this list involved in this research? If so, are these
>     characterizations accurate to your mind?
>  3. There is yet another section where Levine finds some reporting from
>     the early 70s, where NBC News' Rowan Ford conducted a 4 month
>     investigation and found evidence that intelligence files about
>     American anti-war protestors and others had been transferred,
>     perhaps stored, and perhaps processed somehow, over the ARPANET and
>     linked host machines. His report was entered into the Congressional
>     Record as a part of Tunney's hearings in 1975:
>     https://babel.hathitrust.org/cgi/pt?id=mdp.39015078638619;view=1up;seq=7
> 
>     The claim is that these files might have been a part of previous
>     CONUS intel that, in 1972, the Army was ordered to delete. One of
>     the claims in the report is that such files were transferred via the
>     ARPANET to MIT for some reason. Ford had 4 sources for this story
>     who had knowledge of the incident; only one, Richard Ferguson (who
>     apparently was fired from MIT for this disclosure), gave information
>     publicly.
> 
>     Does anyone on this list have knowledge of this incident, and/or
>     whether or not the ARPANET/ARPA IPTO was used to move around,
>     eventually store, or otherwise process these kinds of dossiers?
> 
> 
> These are all the questions I have for now. Thanks for taking the time
> to read.
> 
> 
> -- 
> Eric
> 
> 
> _______
> internet-history mailing list
> internet-history at postel.org
> http://mailman.postel.org/mailman/listinfo/internet-history
> Contact list-owner at postel.org for assistance.
>