Universal Broadband: A 21st Century View

At the risk of feeding an obsession, I’m going to take another stab at explaining what bothers me about President Biden’s universal broadband plan, this time in plain English.

While the goal of providing every American with access to top-quality broadband at a reasonable price is laudable, the plan’s path could hardly be any worse. It has all the features of every engineering project I ever worked on that ended in colossal failure.

Briefly: it misses the moment, it’s naïve, its goals are poorly defined, it has indefensible biases, and it assumes easy things are hard and hard things are easy. The timeline stretches out so far that it can’t be finished during the Biden term and it tries to settle so many scores that it can’t possibly survive the change of administrations that’s bound to take place four years from now, if not sooner.

Missing the Moment

Broadband is included in this omnibus infrastructure plan because people who’ve never cared much about it were forced to take notice  by the pandemic. The rhetoric about broadband being an essential utility – just like electric power! – tells policy people nothing we didn’t already know.

The lesson from the pandemic is that our broadband networks are perfectly fine except for one shortcoming: not enough people are connected. Those who had broadband fared quite well in the working and schooling from home mode, but those who didn’t fell further behind.

This means that we need to get everyone connected at a realistic and adequate level of service as quickly as possible. It does not suggest that we should embark on a five to ten year project to reconnect everyone who’s already connected to some grandiose new network that far exceeds any needs we can identify in the foreseeable future.

Nobody Knows What the Future Will Bring

There’s not a person on the planet who can say with certainly what kind of networks we’re going to need ten years from now, and it’s arrogant to assume otherwise. Present trends only provide vague outlines:

  • Broadband data volume increases
  • Both licensed and unlicensed wireless take on larger shares of the overall load
  • All of the locked down core and middle mile transmission take place over some variety of fiber optic cables
  • Broadband infrastructure require continual maintenance and regular upgrades
  • Networks that connect billions of people are costly but getting cheaper
  • Experts are better at designing and operating networks than are amateurs

Some of the popular applications in 2030 will be new, perhaps the most popular ones. People will insist on better connectivity outdoors for several reasons, not the least of which is that outdoors is a safe place to be in our virus-ridden world.

Infrastructure Plans Must Be Flexible

The administration’s approach to mapping out our broadband future has the aroma of a “People’s Glorious Generational Plan for the Production of Broadband.” This top-down plan is rigid about technology all the way down to the preferred physical plant and the preferred operator.

Its future-focused rhetoric comes from the mid-1990s, when network visionaries failed to predict the rise of the wireless technology that now carries most of our data because they were distracted by a shiny new type of wire and one of many usage scenarios. If Gordon Bell’s vision of symmetrical fiber to every home didn’t come to pass in the 30 years since it was made, why should we believe it ever will?

Against all odds, it probably will because the economics of semiconductor manufacturing favor generality over efficiency. But usage doesn’t follow network design and the presence of symmetrical networks doesn’t dictate symmetrical applications. We’ve had symmetrical LANs, campus networks, and backbones since the 1970s, so the applications would be here already if that were the case.

The role of infrastructure is to support the activities humans wish to do, not to dictate behavior.

Networks of the Future Will be Designed in the Future

The people of 2121 have not time travelled to our primitive age to tell us what kind of network infrastructure to build for them. They will probably have some successor to quantum entanglement that enables them to communicate their thoughts instantaneously in the form they will take after multiple rounds of refinement, debate, and consensus.

We’re in no position to help because their future communication networks will depend on refinements to breakthroughs that haven’t happened yet. It’s unlikely that they’ll be typing on QWERTY keyboards attached to Wintel machines connected to wires hanging from wooden poles and in-ground conduits embedded in interstate highways in 2050.

Even if we had the power to design “future-proof networks” we don’t need to bother. Every element of a every network can and will be improved on a timeline much shorter than those that govern the electric grid, and this goes for wires as well as electronics.

And the grid needs to be more like broadband, BTW.

Political Feasibility

There is a strong bipartisan consensus in Congress for spending on greenfield rural networks and on subsidy programs to connect low-income Americans to broadband networks capable of supporting existing applications. These are clear and defensible needs.

The consensus breaks down when network performance targets are disconnected from reality, technology choices are arbitrary, and suppliers with spotty track records are chosen over those more capable of delivering reliable service. The Biden plan upends the consensus in favor of activist fantasies with no basis in reality.

In the fullness of time, Americans will have the kinds of services the Biden plan seeks to provide. In most markets they will come about organically, as traditional phone and cable companies replace network components – including wires – with better substitutes in their normal upgrade cycles.

Technical Alternatives to Full Fiber

In the harder to serve rural markets, technology plus limited construction subsidies are the answer. These areas are already served by 4G wireless networks, with fixed location 4G and 5G services playing a larger role.

Elon Musk’s Starlink service is another emerging option that doesn’t require fiber to the farm. Both terrestrial wireless and Starlink require base stations connected to fiber backhaul.

Backhaul buildouts are a necessary step to full fiber deployment. The backhaul buildouts will only be helpful to full fiber in the future if they’re accessible to firms with an interest in taking on the last part of the hookup, of course.

A Role for Government

There could be a government role in ensuring middle mile accessibility, but I’m uncertain how it would work in detail. An open middle mile is automatic when the backhaul is built by the independent fiber operators because it’s in their interest to sell to as many customers as possible.

Government has a variety of tools at its disposal to accelerate rural fiber deployment, from subsidies to mandates. These tools should be applied to the middle mile first.

There may never be a real market for full fiber buildouts in farm country, but there is already demand for backhaul. So the government role is probably making the things that are already happening happen faster.

Rural networking is not a blank slate.

The Non-Profit Delusion

The Biden plan explicitly favors non-profit providers such as rural electric co-ops over commercial ones. The assumption is that non-profits can offer lower consumer prices.

While this idea has naïve appeal, there are some downsides. Research is mixed at best on whether subscriptions from non-profit players are cheaper than those from for-profit firms when all costs, subsidies, and transfers are considered. No for-profit ISP charges a connection fee that rivals the $3,000-3,500 charge levied by the city of Ammon, Idaho.

The data compiled by fans of the non-profit model such as Chao and Park of OTI doesn’t confirm the claim that non-profit broadband is cheaper:

In all, the statistical evidence comparing prices for like services between municipal and private providers within the same market mostly indicates that private providers charge equal prices on average. This result is expected by the law of one price. Comparing prices across markets with and without a government-run networks, however, reveals that prices are higher in markets with government-run networks. There is no reliable evidence, either for within- or between-market comparisons, that government-run networks lower prices.

Yet even the White House parrots the Chao and Park report.

Broadband Equity: Addressing Disparities in Access and Affordability

Yesterday’s hearing on the Biden broadband plan in the House Communications and Technology subcommittee failed to reveal evidence that the US government needs to undertake a ten year project to build one big symmetrical wired network.

To the contrary, witnesses agreed that the urgent needs are immediate subsidies for greenfield builds and low-income support to access existing networks. With the emergence of technical alternatives to wired broadband such as 5G, 6G, and low earth orbit satellite constellations, we can expect fiber to revert to its traditional role as a backhaul technology.

At the same time, telephone twisted pair copper wire will continue to be replaced by fiber in neighborhoods, as needed, at a reasonable pace. The way forward is to prioritize urgent needs over long term visions.

In cases where a new wireline network is the only solution that will get a rural community online, of course that network needs to be all fiber and potentially symmetrical. But such cases are rare.

All in all, practicality needs to trump 1990s visions of future networks that never came to be. Building new networks in urban areas does not get more people online.