Much of this was Unnecessary
You don't need great protocols for a terrible Internet
|Nov 25, 2020|
Soon, I will be punished. It will probably happen tonight. I’ll try to talk to my family—you know what that means, see? You know it means the Internet. The Internet is part of the fabric of everyday life; we can’t exist in this form without it. So I’ll use it, because that’s normal, as we agreed, to talk to my family. You know I have to use it because of the pandemic. We all know we cut ourselves away from the little social topologies that helped make a lot of things out of core functions worth it.
This is where the punishment starts. Captchas, dropped calls, broken websites, inexplicably failed logins, all so many reminders that it’s wrong to expect the most basic privacy. I mean just tapping the brakes on a gang of sociopath varsity engineers demanding that, in exchange for basic social life, I provide them with a list of my fears. That’s the fairly complete and very predictive psychometric profile that is extracted from every website, keystroke, pause, and deleted word. So that for the rest of my life I can face a minor psychic disturbance whenever someone decides it’s time to try to sell me something. When the smartest people and fastest computers from this side of the world will do their best to take what they know about me and use it to make me feel smaller, to make me feel just a bit broken, to make me vulnerable to the solution—that’s probably something carcinogenic or voting for the new variant of far right populism they put together.
Cool, cool. That’s all fine. Standard capitalism hostage situation. But I want to point something out. The distributed topology? That soundbite that media theorists and network philosophers or whatever have been worshipping for more than a decade straight? How it isn’t centralized, how it routes around censorship as damage, how it’s automatic and decentralized and the engine of a new era? Unnecessary. We don’t need it.
Does the Internet feel particularly decentralized to you? Is it that the vast majority of your traffic goes to just a few sites? Or that that traffic is carried on infrastructure owned by those sites? Is it the photos of the data centers and massive hubs? No, this just feels like everything else. It feels like where everything else landed, too. What we got after gullible people listed the ways that, in theory, the technology could create some kind of informational utopia: telegraph, wireless telegraphy (radio), telephone, television. They were right. The basic specification allows for a utopian society or a society trending toward utopia built atop a telegraph or whatever else. It doesn't really matter, though. Each centralized immediately with the certainty that you’d get from a time-traveller sent from the future who knows what’s going to happen. Imagine the drama. It has happened. It will happen, they exclaim. Here’s a trick: go out and find a decent map of the Internet that shows a distributed topology. Psych! There aren’t any, because the Internet has never been distributed.
The Internet could be distributed, but capitalism removed any chance of that. John Day, the most interesting (and entertaining) network scientist alive today, claims that in its evolution the Internet “lost a layer” because the Internet was supposed to exist on top of local networks of every which type. The original idea is that you’d have all these governments, communities, colleges building their own networks—this idea was held by the same frequently-blindsided people that expected the orderly transition from public ownership—and those networks would remain unique to those groups’ needs. They’d enforce their own surveillance, privacy, addressing, routing etc policies, like free peoples should. You can see how this would work. Then they would install protocols that would let them, insofar as they saw fit, make global connections. Instead, Day observes, the Internet Protocol (IP) pushed by DARPA took over those large component networks, too. It wasn’t really an inter-network, so much as a single global network with subdivisions. The fact that ethernet dominated small, local networks meant that there would be little heterogeneity there, either. I’ll add that all of this makes surveillance a lot easier, too.
(Industry people: by heterogeneity I don’t mean a bunch of MPLS, SDN, or whatever disasters, I mean architectures that differ from the bottom up. Overloading the application layer doesn’t count!)
But it wasn’t the protocol. It wasn’t IP. The Internet Protocol is still fine, as far as it goes. Go read the spec: it can still work with a radically heterogeneous inter-network. It was designed to layer on top of distinct local networks, and it still can.
It’s just that those networks never appeared. We never built them. Not at any scale, anyway. After a brief spasm of heterogeneity, the networking industry got down to its destiny and proceeded toward oligopoly and centralization. I doubt the first generation of nerd engineers who formed those companies even wanted to go that corporate route, but they never had a choice. We see firms and markets for packet-switched products in the 1970s, at very small scale, and by the late 1980s, the industry was congealing and had already ossified around two mainline protocol implementations: IP and ethernet. And that’s what we’ve used since. No need for all that protocol agnosticism, all that flexibility. It’s there waiting for us in the postcards from the 60s, the series of TCP and TCP/IP specifications that seem at least to be innocent of that whole it has happened, it will happen thing.
So, to take a step back (I can never edit on substack or I will never produce anything and/or it will consume my days), the reason why the Internet isn’t quite an Internet according to John Day is that the networks collapsed into the internet. But it wasn’t a DARPA conspiracy (Day doesn’t like DARPA much). It was capitalism. Just like it is capitalism that structured and controls our platforms, our YouTubes and Facebooks.
Go back to the IP spec and you’ll see preparation for a lot of nodes exchanging information. Also not what we’re doing. A few firms control most of the traffic that goes over infrastructure they own too. (Side point: the entire net neutrality project assumes a separation between host/platform and network that no longer exists.) The data is stored and the remote computation is accomplished in a small number of places. We don’t need a distributed and interoperable architecture for that. When was the last time you required a globally distributed, interoperable, and content-agnostic architecture? Nowadays we’re getting better at simulating those in massive centralized data center. That’s some especially computational ideology. And when the humiliations come for you this holiday, just remember, it isn’t the Internet. The Internet Protocol spec doesn’t have anything against you. But something does.