“designed from the beginning to operate while in tatters”

August 3, 2024 4:47 PM

Was the Internet Designed to Survive a Nuclear Attack? is an essay by computer historian Chris McKenzie which traces the origins of the popular myth that the Internet was the result of an attempt to create a military command and control network that wasn’t vulnerable to a single nuclear strike. Via Bruce Sterling, who wrote an early version of the narrative.
posted by Kattullus (14 comments total)

internet archive says this is from late 2022, juuust before the ChatGPT release . . . using regex for searching text is charming (as is its mention of maybe using BERT for this)

Checking my college transcript, I see I took CS112 (Network Modeling Fundamentals aka "Queing Theory"), eg:

$P(X=k)=\frac{e^{-\lambda }{\lambda }^k}{k!}$  [1]

with Dr Kleinrock in 1990 . . . little did I know then what was coming, though I did find the national NSF network diagrams (featured in this story) posted in the CS floor's display boards mildly interesting.

Kleinrock was whip-smart and he led a tight class. (I also met the founders of what became Blizzard in that class so it was a pretty formational experience for me LOL) . . . the 90s were a very fun decade for CS people, the world was our oyster . . .
posted by torokunai at 5:31 PM on August 3 [6 favorites]

"How it was sold to the military that initially funded it" might not be the same as "what it was designed to do", just putting that out there.
posted by mhoye at 5:41 PM on August 3 [13 favorites]

yeah Dr Jefferson was working on Time Warp when I took his class, CS 111 (Operating Systems) in '88. This was funded by DARPA so had to simulate a tank battle against Russians IIRC. Disturbingly relevant in 2024 alas.
posted by torokunai at 5:45 PM on August 3 [1 favorite]

This type of obsessively detailed and cited amateur investigation is one of my favorite things about the web. It's also fascinating how our standards of attribution have changed over the last 50 years. The idiom of the "backhoe threat of the internet" is exactly the type of thing I'd expect to hear a dozen times in classroom lectures or conference presentations without necessarily seeing it written in an enduring document. But over the last 10 years "Know Your Meme" has made it somewhat de rigueur to find the exact coinage.
posted by midmarch snowman at 7:01 PM on August 3 [3 favorites]

haha what

I guess I brought this to Bruce Sterling's attention? I'm pretty surprised his accounts at Twitter and The WELL didn't come up in research.
posted by Pronoiac at 9:20 PM on August 3 [5 favorites]

There is maybe a related study that could be done on those languages where it is “the Internet” and those, like French - who normally like to assign everything a gender- where remains simply “Internet”. We never talk about “the Usenet” and “INTERNET” alone sounds much more like what a secret military project would call the thing.
posted by rongorongo at 2:35 AM on August 4

Good to know - I’ve heard (the) Internet cited many times as the kind of incidental benefit that justifies a colossal defence research budget.
posted by Phanx at 2:47 AM on August 4

Bloody revisionists. Next they'll be telling us that the invention of Teflon had nothing to do with making it easier for the Apollo crews to wash up their frying pans in zero G.
posted by flabdablet at 3:26 AM on August 4 [3 favorites]

The sidebar definition of "common knowledge" is fascinating, and one which I'll probably reference for years to come.
posted by cheshyre at 5:57 AM on August 4 [2 favorites]

The sidebar definition of "common knowledge" is fascinating, and one which I'll probably reference for years to come.

Call me elitist, but these days any time I see the word “common” used in this way, it does not evoke the sort of folksy epistemic unassailability the idiom typically implies, but rather “common” as in “of the common people,” opposing terms like “rarefied,” “specialized” or “educated.” Calling people “common” is a pretty shitty way to think of their essential worth as people, but an entirely valid critique of their innate credibility on important and nuanced matters. We have an unfortunate tendency to confuse those two things.

Example: The claim that there are exactly two well-defined biological sexes in human beings entailing intrinsic behavioral roles is indeed common knowledge, and that’s exactly what’s wrong with it.

Maybe in the modern day we should substitute “common” with “basic” (omitting the gendered alliteration that often follows, of course), or if we’re feeling really generationally adventurous, “mid.”
posted by gelfin at 8:01 AM on August 4

rare example of a true, modern, functional anarchy 🤘
posted by HearHere at 10:01 AM on August 4 [2 favorites]

What a tremendous article. Thank you so much for this post!!
posted by riverlife at 3:25 PM on August 4 [1 favorite]

Fantastic article! I'm not sure it really solves anything - we continue not to know exactly how or why all this nonsense really started. But it doesn't really matter.

As I started reading this and pondering over the idea of a network that could sustain massive damage at any point and continue to operate, it occurred to me that we clearly have that in terms of the physical network. However, we have many single points of failure for the systems that run on that network, some well-known like AWS, some that most had never heard of until they were suddenly on everyone's lips like CrowdStrike. Over decades, we've built, rebuilt, improved and expanded the network and made it more robust physically. But we've put all our faith in actually being able to communicate and do things on that network in the hands of a few companies.

This doesn't seem like a good idea.
posted by dg at 4:50 PM on August 4 [2 favorites]

pondering over the idea of a network that could sustain massive damage at any point and continue to operate, it occurred to me that we clearly have that in terms of the physical network.

Maybe not as much as is generally believed. TCP is an unbelievably shitty transport layer protocol, not even vaguely resilient in the face of packet loss caused by anything but simple congestion. If more than about 10% of packets fail to get through because e.g. the last-mile link is noisy, TCP's congestion-control backoff becomes totally maladaptive and it just keeps on backing off until there's virtually no usable connectivity to be had. On the flip side, if you tunnel one TCP connection inside another one, the two layers can get into a fight over retrying timed-out acknowledgements to the point that almost all of what the outer layer is carrying is completely superfluous inner-layer retransmissions.

It never had to be that way. Forward error correction has been a thing for longer than the ARPANet has, but the design of the entire TCP/IP stack completely ignored it and all of TCP's resulting deficiencies are still being suffered today as a result.

The Internet is built with a certain amount of redundancy in its major backbone links, but taking out enough of those to stop it from functioning to the extent that we now routinely rely on it to would be much less hard than most people seem to think.

Starlink is a great big space-based mesh network that doesn't rely on TCP/IP to implement that mesh, but it's in private hands, so any technical resiliency it offers is more than offset by its political brittleness.
posted by flabdablet at 1:39 AM on August 5