As I pointed out in the last post, the Engineers Letter EFF filed with the FCC in the Internet Freedom docket is riddled with errors. It attacks the Commission’s understanding of the Internet with a “Brief Introduction to the Internet” originally filed with the DC Circuit in the challenge to the 2015 Open Internet order. An attack on today’s FCC written for an amicus brief isn’t really on point.

Chairman Pai could very well “lack lack a fundamental understanding of what the Internet’s technology promises to provide, how the Internet actually works, which entities in the Internet ecosystem provide which services, and what the similarities and differences are between the Internet and other telecommunications systems the  FCC regulates as telecommunications services” as EFF claims. But trying to prove that with a letter written for an entirely different purpose is disingenuous at best.

In fact, the EFF letter is guilty of all the charges it raises against today’s FCC.

What Internet Technology Promises to Provide

Any claim about the “promise of the Internet” is necessarily subjective. Some people claim the Internet promises free access to information, the freedom to communicate, and democracy; but in many cases it brings about controlled speech and surveillance. While the negatives are commonly associated with authoritarian governments such as China and Saudi Arabia, this isn’t the whole story.

The financial model for most of the large “Edge Service” networks is based on advertising; this involves tracking visits to websites and the creation of dossiers of the interests and activities of individual users. The EFF doesn’t mention any of this, but surveillance states are jealous of the dossiers Google and Facebook compile not just on visitors to their sites but on visitors to the websites that use their tracking code.

In reality, the Internet promises to connect every network that wants to participate on terms of each network’s choosing. When we get to specific terms and conditions of the interconnection, we’re in the realm of local and national policy. The design of the Internet doesn’t specify what we can do with our connections.

The claim that Internet inherently requires “openness and non-Interference” is wishful thinking. We Americans want it to be open and indiscriminate, but there is nothing in the design of the Internet that forces it to behave that way. And the idea that designers of a system are free to create the public policy for the system is indeed very troubling. Would you extend that privilege to firearms designers?

The Great Firewall of China proves that the Internet behaves the way lawmakers want it to behave. And that’s why policy discussions are more important that technical claims.

The Internet simply provides us with an opportunity to communicate; what we do with it is up to us.

How the Internet Actually Works

The Internet is simply a communication system that interconnects networks of computers to other networks of computers. This interconnection has two parts: 1) A physical interconnection with a common technology (most commonly Ethernet in addition to TCP/IP); and 2) An agreement to interconnect, either directly or through a third party.

Third party, for fee “transit” networks, which the EFF letter fails to mention, are the most common mode of interconnection for the small networks, and settlement-free peering is the common interconnection for the small number of extremely large, dominant networks such as Google, Facebook, Comcast, Verizon, and AT&T.

Ordinary businesses and regional ISPs connect to the Internet through transit providers such as Level 3. Level 3 provides both physical connections to the entire Internet and agreements with all of its constituent networks, either directly or through other transit networks. So most networks connect to each other under agreements by service providers.

Which Players in the Internet Ecosystem Provide Which Services

The Internet consists of connections, agreements, and information systems. There are no bright lines in the Internet segregating these three things.

The large edge services – the monopolies enjoyed by Google, Facebook, et al. – have networks with global reach. They have facilities all across the US and in other nations. But so do the transit networks and the “backbone networks” owned by large ISPs. So everybody is in the connection business.

In addition to the basic interconnection agreements already mentioned, networks also agree to share routing information with each other via common routing information protocols like BGP. Networks also agree to share domain information with a DNS, a non-routing system used by large networks to organize their resources.

DNS also provides lookups between names and numbers, like the telephone directory. But unlike the telephone directory, DNS is not managed solely by connection merchants. It’s a shared database updated in close to real time by domain owners. Access to an Internet-based service can require dozens of DNS database queries in sub-second intervals of time.

DNS is a General-Purpose Database

So DNS is much more than a simple directory. EFF claims DNS’s “primary  purpose is to enhance functionality otherwise provided by the internetwork layer”. This is such a vague statement as to be meaningless. We can also claim that DNS “enhances functionality otherwise provided by the application layer”.

EFF also claims that “DNS allows a vital level of abstraction” because it makes it unnecessary to remember IP addresses and allows IP addresses assigned to specific computers to change over time.  But the Internet had means of doing this before we had DNS: we used to download a “hostfile” every week that had computer names and associated IP addresses in it. You can use a hostfile today to bypass DNS if you want.

The benefit of DNS over the hostfile is the same as for any database over a text file equivalent: concurrent access by multiple users, resilience, and instant updates. Having a shared database that spans the entire Internet – and allows for the storage and retrieval of any information, not just domain names and IP addresses – provides a number of benefits.

Service networks can use DNS to optimize resources, balance load, and match up content formats with users. DNS supports the application layer, not just the internetwork layer. It’s completely misleading to claim DNS is merely a tool to make packet routing happen.

Arguably, DNS has nothing at all to do with routing. Routing is the job of IP and BGP.

The Similarities and Differences between the Internet and the Telephone Network

The worst part of the EFF letter is the section comparing the Internet to the telephone network. I’ll write a whole post on the issues with that section.