Why Dumb Networks Are Better
First published in FEE, in March 2015.
In computer and communications networks, decentralization leads to faster innovation, greater openness, and lower cost. Decentralization creates the conditions for competition and diversity in the services the network provides.
But how can you tell if a network is decentralized, and what makes it more likely to be decentralized? Network “intelligence” is the characteristic that differentiates centralized from decentralized networks — but in a way that is surprising and counterintuitive.
Some networks are “smart.” They offer sophisticated services that can be delivered to very simple end-user devices on the “edge” of the network. Other networks are “dumb” — they offer only a very basic service and require that the end-user devices are intelligent. What’s smart about dumb networks is that they push innovation to the edge, giving end-users control over the pace and direction of innovation. Simplicity at the center allows for complexity at the edge, which fosters the vast decentralization of services.
Surprisingly, then, “dumb” networks are the smart choice for innovation and freedom.
The telephone network used to be a smart network supporting dumb devices (telephones). All the intelligence in the telephone network and all the services were contained in the phone company’s switching buildings. The telephone on the consumer’s kitchen table was little more than a speaker and a microphone. Even the most advanced touch-tone telephones were still pretty simple devices, depending entirely on the network services they could “request” through beeping the right tones.
In a smart network like that, there is no room for innovation at the edge. Sure, you can make a phone look like a cheeseburger or a banana, but you can’t change the services it offers. The services depend entirely on the central switches owned by the phone company. Centralized innovation means slow innovation. It also means innovation directed by the goals of a single company. As a result, anything that doesn’t seem to fit the vision of the company that owns the network is rejected or even actively fought.
In fact, until 1968, AT&T restricted the devices allowed on the network to a handful of approved devices. In 1968, in a landmark decision, the FCC ruled in favor of the Carterfone, an acoustic coupler device for connecting two-way radios to telephones, opening the door for any consumer device that didn’t “cause harm to the system.”
That ruling paved the way for the answering machine, the fax machine, and the modem. But even with the ability to connect smarter devices to the edge, it wasn’t until the modem that innovation really accelerated. The modem represented a complete inversion of the architecture: all the intelligence was moved to the edge, and the phone network was used only as an underlying “dumb” network to carry the data.
Did the telecommunications companies welcome this development? Of course not! They fought it for nearly a decade, using regulation, lobbying, and legal threats against the new competition. In some countries, modem calls across international lines were automatically disconnected to prevent competition in the lucrative long-distance market. In the end, the Internet won. Now, almost the entire phone network runs as an app on top of the Internet.
The Internet is a dumb network, which is its defining and most valuable feature. The Internet’s protocol (transmission control protocol/Internet protocol, or TCP/IP) doesn’t offer “services.” It doesn’t make decisions about content. It doesn’t distinguish between photos and text, video and audio. It doesn’t have a list of approved applications. It doesn’t even distinguish between client and server, user and host, or individual versus corporation. Every IP address is an equal peer.
TCP/IP acts as an efficient pipeline, moving data from one point to another. Over time, it has had some minor adjustments to offer some differentiated “quality of service” capabilities, but other than that, it remains, for the most part, a dumb data pipeline. Almost all the intelligence is on the edge — all the services, all the applications are created on the edge-devices. Creating a new application does not involve changing the network. The Web, voice, video, and social media were all created as applications on the edge without any need to modify the Internet protocol.
So the dumb network becomes a platform for independent innovation, without permission, at the edge. The result is an incredible range of innovations, carried out at an even more incredible pace. People interested in even the tiniest of niche applications can create them on the edge. Applications that only have two participants only need two devices to support them, and they can run on the Internet. Contrast that to the telephone network where a new “service,” like caller ID, had to be built and deployed on every company switch, incurring maintenance cost for every subscriber. So only the most popular, profitable, and widely used services got deployed.
The financial services industry is built on top of many highly specialized and service-specific networks. Most of these are layered atop the Internet, but they are architected as closed, centralized, and “smart” networks with limited intelligence on the edge.
Take, for example, the Society for Worldwide Interbank Financial Telecommunication (SWIFT), the international wire transfer network. The consortium behind SWIFT has built a closed network of member banks that offers specific services: secure messages, mostly payment orders. Only banks can be members, and the network services are highly centralized.
The SWIFT network is just one of dozens of single-purpose, tightly controlled, and closed networks offered to financial services companies such as banks, brokerage firms, and exchanges. All these networks mediate the services by interposing the service provider between the “users,” and they allow minimal innovation or differentiation at the edge — that is, they are smart networks serving mostly dumb devices.
Bitcoin is the Internet of money. It offers a basic dumb network that connects peers from anywhere in the world. The bitcoin network itself does not define any financial services or applications. It doesn’t require membership registration or identification. It doesn’t control the types of devices or applications that can live on its edge. Bitcoin offers one service: securely time-stamped scripted transactions. Everything else is built on the edge-devices as an application. Bitcoin allows any application to be developed independently, without permission, on the edge of the network. A developer can create a new application using the transactional service as a platform and deploy it on any device. Even niche applications with few users — applications never envisioned by the bitcoin protocol creator — can be built and deployed.
Almost any network architecture can be inverted. You can build a closed network on top of an open network or vice versa, although it is easier to centralize than to decentralize. The modem inverted the phone network, giving us the Internet. The banks have built closed network systems on top of the decentralized Internet. Now bitcoin provides an open network platform for financial services on top of the open and decentralized Internet. The financial services built on top of bitcoin are themselves open because they are not “services” delivered by the network; they are “apps” running on top of the network. This arrangement opens a market for applications, putting the end user in a position of power to choose the right application without restrictions.
What happens when an industry transitions from using one or more “smart” and centralized networks to using a common, decentralized, open, and dumb network? A tsunami of innovation that was pent up for decades is suddenly released. All the applications that could never get permission in the closed network can now be developed and deployed without permission. At first, this change involves reinventing the previously centralized services with new and open decentralized alternatives. We saw that with the Internet, as traditional telecommunications services were reinvented with email, instant messaging, and video calls.
This first wave is also characterized by disintermediation — the removal of entire layers of intermediaries who are no longer necessary. With the Internet, this meant replacing brokers, classified ads publishers, real estate agents, car salespeople, and many others with search engines and online direct markets. In the financial industry, bitcoin will create a similar wave of disintermediation by making clearinghouses, exchanges, and wire transfer services obsolete. The big difference is that some of these disintermediated layers are multibillion dollar industries that are no longer needed.
Beyond the first wave of innovation, which simply replaces existing services, is another wave that begins to build the applications that were impossible with the previous centralized network. The second wave doesn’t just create applications that compare to existing services; it spawns new industries on the basis of applications that were previously too expensive or too difficult to scale. By eliminating friction in payments, bitcoin doesn’t just make better payments; it introduces market mechanisms and price discovery to economic activities that were too small or inefficient under the previous cost structure.
We used to think “smart” networks would deliver the most value, but making the network “dumb” enabled a massive wave of innovation. Intelligence at the edge brings choice, freedom, and experimentation without permission. In networks, “dumb” is better.
My O’Reilly conference talk on this topic: