[ih] Fwd: Re: When the words Internet was design to survive a nuclear war appeared for the first time?
Steve Crocker
steve at shinkuro.com
Sat Feb 16 08:25:05 PST 2019
My brother forwarded the appended thread. It reminded me that I had
previously been unsuccessful at joining this list, but I am now on the list.
Over the years I have heard many times from many different people and also
seen in print the idea that the Arpanet was motivated and built for nuclear
survivability. Moreover, Steve Lukasik, the director of (D)ARPA for
several of the critical years, included this as one of his reasons for
signing the checks for the Arpanet.
I think it will be helpful to make a couple of distinctions regarding
network survivability. Let me offer two dimensions and at least two levels
of disruption for each.
*Equipment outage*: When a link or a router becomes unusable, it's
necessary to route around the loss. It makes a big difference if only a
few links or routers are down versus a significant fraction are down. In
normal circumstances, it is expected some lines and some routers will be
out of service from time to time. In contrast, if there is an attack, the
outages might be substantial and perhaps purposefully coordinated.
*Traffic level*: In normal operation, the total traffic is within the
capacity of the network. In extraordinary times, the traffic level surges
beyond the capacity of the system. Surges happen for various reasons.
There might be a very popular web site, e.g. the Victoria Secrets incident,
or a DDoS attack.
Arguably the Arpanet design addressed survivability for the case when a
small number of links or routers were out and traffic levels were normal.
Alex colorfully describes Frank Heart's concerns for the risks to
individual IMPs placed in the very risky environments of universities.
The Arpanet design did not address the issue of large scale outages nor did
it include strategies for dealing with overload. In contrast, a serious
design to address post-nuclear operation would have had to address the
combination of substantial loss of equipment and a huge spike in urgent
traffic.
In the conversations I've had with Lukasik on this matter, he says he had
in mind the technology could lead to designs that would have that level of
survivability, and that it was helpful to include this in defending the
funding for the network. And, as Vint says, subsequent projects explored
reconstitution in the event of certain kinds of disconnections. However, I
believe the level of outages explored in those projects were well below the
levels that would have occurred in a large scale nuclear confrontation.
In 2017, I ran this question past Larry Roberts. His short reply was he
would have had to connect each IMP to four others instead of only two or
three others.
Steve
-------- Forwarded Message --------
Subject: Re: [ih] When the words Internet was design to survive a
nuclear war appeared for the first time?
Date: Thu, 14 Feb 2019 12:30:36 -0500
From: Vint Cerf <vint at google.com>
To: Alex McKenzie <aamsendonly396 at gmail.com>
CC: internet history <internet-history at postel.org>
Alex is essentially correct. Paul Baran's work WAS aimed at post-nuclear
survival but he never got to try his ideas out as they
were rejected as unimplementable or uninteresting by a circuit-switching
oriented Defense Communications Agency.
Larry Roberts was clear that the ARPANET was intended to support
resource sharing.
By the time Bob Kahn and I started working on Internet with its focus on
command/control, the issue of survivability
was back on the table. The multi-network design contemplated multiple
networks operated by distinct entities
(in the DoD perspective it was multiple countries or aggregates like
NATO) and resilience was important. I went
so far as to commission a test in which we flew packet radios in
Strategic Air Command aircraft, artificially "broke"
the ARPANET up into fragments and re-integrated them through ground to
air packet radio connectivity. I was
particularly worried about the partitioning of a constituent network
which would cause great confusion for the
routing algorithm (a source gateway might know know to which "half" of
the fragmented network a packet should
be sent. My hazy recollection is that Radia Perlman came up with a way
to solve that problem that involved
creating new autonomous systems out of each "piece" and re-enabling
routing algorithms.
vint
On Thu, Feb 14, 2019 at 12:18 PM Alex McKenzie <aamsendonly396 at gmail.com
<mailto:aamsendonly396 at gmail.com>> wrote:
Miles,
I believe the emphasis on survivability came from Frank Heart.
Building the early ARPAnet was a very risky project, in the sense
that there was a tight deadline, it would be easy to see if it
worked or not, and most people didn't believe it would work.
Frank's reputation was very much on the line. The ruggedized IMP
cabinet was part of his emphasis on controlling everything the team
could control, to minimize risk. But the particular risks the
ruggedized cabinet was intended to protect against were:
- careless site personnel, who cared about their own computers but
might be expected to stick the IMP in a storage closet where
maintenance workers would bump it, and
- graduate students who might be inclined to study it, perhaps with
destructive results. (Lest this seem outlandish, the TIP in Hawaii
was a sore spot of unreliability when I was running the NCC -
turned out a graduate student was crashing it every day by taping
into its power supply which was just right for his project. The TIP
was NOT in a ruggedized box.)
The group was not trying to protect against EMP.
More generally, if the ARPAnet had been designed to survive a
nuclear attack it would have been necessary to insure that the
IMP-to_IMP circuits did not go through the small number of Telco
offices which made up the Telco backbone. No effort was made to
influence the provisioning of these circuits, and it can be presumed
that loss of only a few major cities would have resulted in most of
the leased lines disappearing.
Cheers,
Alex
On Thu, Feb 14, 2019 at 10:48 AM Miles Fidelman
<mfidelman at meetinghouse.net <mailto:mfidelman at meetinghouse.net>> wrote:
Bernie,
On 2/14/19 9:28 AM, Bernie Cosell wrote:
> On February 14, 2019 09:13:42 Alejandro Acosta
> <alejandroacostaalamo at gmail.com>
> <mailto:alejandroacostaalamo at gmail.com> wrote:
>
>>
>> Today I was reading some news about Internet and in one of
>> them I
>> found the phrase (that all of you have listened before):
>> "Internet
>> (ARPANET) was intended to survive a nuclear war", however, as
>> far as I
>> know, this is kind of a myth, right?, ARPANET was intended as
>> a research
>> network and the "war" part if very far away from the thuth.
>
> my take on that is that there were two lines of thought
> leading up to the
> ARPAnet. very very roughly: one was paul baran's, who was
> thinking
> about how the military command and control might be able to
> continue functioning in the event of an attack, and JCR
> Licklider, who was thinking
> about how wide-spread researchers could share resources, ideas
> and results
> to better collaborate.
>
> when the ARPAnet got funded by the DoD, Baran's story was the
> easier to
> understand to the average person, raather than the more
> diaphanous idea
> of researcher collaboration. so Baran's take kinda caught the
> public
> imagination, but the reality for those of us working on it was
> the it was
> {somehow :o)} to be a research tool.
>
You were involved a lot earlier than I was. Perhaps you could
comment on how much folks thought about fault-tolerance in the
early days. It's always struck me that things like
continuity-of-operations, in the face of node & link outages,
and no-single-point-of-failure, were baked in from the
beginning. You know - all the stuff that would allow the net to
survive everything from backhoes to natural disasters, and
coincidentally, nuclear war.
On the physical side, the early IMPs were pretty rugged boxes
(not so much C/30s and such). Were any of the IMPs built to
withstand EMP?
Miles
-- In theory, there is no difference between theory and
practice.
In practice, there is. .... Yogi Berra
_______
internet-history mailing list
internet-history at postel.org <mailto:internet-history at postel.org>
http://mailman.postel.org/mailman/listinfo/internet-history
Contact list-owner at postel.org <mailto:list-owner at postel.org> for
assistance.
_______
internet-history mailing list
internet-history at postel.org <mailto:internet-history at postel.org>
http://mailman.postel.org/mailman/listinfo/internet-history
Contact list-owner at postel.org <mailto:list-owner at postel.org> for
assistance.
--
New postal address:
Google
1875 Explorer Street, 10th Floor
Reston, VA 20190
--
Dave Crocker
Brandenburg InternetWorking
bbiw.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20190216/146c4e3e/attachment.htm>
-------------- next part --------------
_______
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.
More information about the Internet-history
mailing list