Is The Internet Getting Too Much Like the Phone Network?

My brilliant friend Bob Frankston, implementer of VisiCalc and a prime mover behind much home networking technology, thinks that we’re moving in the wrong technical and business direction in most of our attempts to extend residential broadband. Bob believes that we are being pushed by what he calls the Regulatorium into making the Internet look more like the phone network to the detriment of almost everyone except the phone companies who will still have a business model.

Bob may well be right.

The primary ways we Americans get broadband Internet access are through cable or telephone company connections. We pay relatively high rates for relatively slow connections if we compare ourselves to much of the developed world. Places with better deals for access often have either some publicly-owned infrastructure or strict rules which require that the telcos and cablecos who got under the streets in the days of monopoly franchises open “their” pipes to competing services.

Bob is actually advocating not only a different ownership structure but also a different architecture for Internet access and Internet carriage than is being implemented in much of the world today.  He argues persuasively (and passionately) that the Internet – whose strength has always been its decentralization – is becoming too centralized and hierarchical just at the time when new technology makes further flattening both possible and desirable.

Today’s Internet is a distributed hierarchy. The devices in my house submit to the central control of my router.  My router forwards all its packets through a single pipe (a T-1 at this location; a radio connection at our summer camp) to an ISP-owned router.  My ISPs are small and local but most people are talking directly to the router of their phone or cable company.  Typically the ISP router  is connected to several large pipes which give it access to “the Internet” – that is, to everybody else and all the rest of the content on the Internet. At the other end of these pipes are typically even larger routers – almost all of these belong to major carriers like at&t, Verizon, Cable & Wireless etc. These networks connect in gargantuan routers either at private “peering points” between networks or in quasi-public “meet me” rooms.

Although some packets of data on the Internet stay local, most travel up and down the hierarchy from source to destination and answering packets travel back – not necessarily by the same route but usually through an equally hierarchical path.  The strength and durability of the Internet comes from the fact that there is no central control of the path that packets take – routers sort of ad lib connections as they go along.  So failures are usually routed around, perhaps at the expense of some congestion.  Moreover, the protocols used for most Internet traffic can tolerate some packet loss so the system doesn’t have to be perfect.  Not having to be near perfect (five nines of reliability in telco jargon) is a huge engineering advantage – one of the reasons why the Internet is so much cheaper to use than the networks that went before it.

Bob Frankston questions whether this hierarchical routing is needed at all. We are very close to being able to provide local connectivity through a mesh of radio antennas. Although current implementations rely on a two-level hierarchy of access devices (computers or, increasingly, phones) and access points, the engineering of mesh of all peers is well understood. In such a mesh, each computer is aware of its neighbors and devotes a small part of its computer and radio power to passing on packets from one neighbor towards another. Remember, even in today’s Internet the whole path a packet takes from origin to destination is not predetermined.

Super peers or higher capacity connections might be desirable on heavily traveled routes (even this isn’t a given since a highly traveled route will tend to be well-populated with computers which can serve as relays).  Special connections are certainly required to cross uninhabited areas, oceans, and radio-unfriendly terrain. Bob thinks of this as all being like the network of local roads, regional roads, and super highways that we’re all familiar with. 

This analogy is important to him.  He describes today’s broadband implementation as “running trolley tracks to your door” rather than giving your driveway access to the road network. Reasoning further from the analogy, he points out that the road system would be impossibly complex if we had to pay a toll each time we left the driveway. (As an exTransportation Secretary I’d argue that we do pay gas tax but his point is well taken).

In fact, says Bob, the only reason for building a complex hierarchical system is to create billable events for the owners of that system. But all the users of the system are unduly constrained by the complexity. It’s a good argument.  I hope Bob’ll correct me if I got parts wrong.

Open spectrum is necessary to building a flatter meshed Internet (my opinion, not necessarily Bob’s).

Yochai Benkler of Yale has explained how it is that each of us can afford to be a relay in a volunteer meshed network without undue expense.