What might we do with the Internet of the future?

The world is about to be filled with autonomous cars, social robots, and delivery drones. These will all need to communicate and collaborate.

We would like to re-create the empathy and presence of in-person meetings through virtual worlds. This places new demands on the network, with responses that must fit within the limits of human perception.

Vast arrays of sensors tied to decision engines continually tweak the flow of resources in our homes, cities and factories. These need to be coordinated in real time, and protected from sabotage.

Our economy, transport, health, and culture is increasingly dependent on the Internet as infrastructure. As we do radically different things to the past, we need something greatly improved.

We’ve been through this transition before with previous industrial revolutions. Commuting to work requires cheap and reliable cars. Such new mass behaviours took half a century to emerge after the basic transport technologies first became available.

We are now at that half century since the start of packet networks. We need a cheaper and more reliable digital infrastructure that can accommodate all these new uses, and more.

The Internet falls short
We can’t afford to build a bespoke infrastructure for every application. So can the Internet we have today meet these future needs as a general-purpose shared infrastructure?

Definitely not. It falls short on many accounts: It lacks the performance, security, reliability, and cost structure needed. Why so many problems?

The Internet does one thing exceedingly well: it provides simple, ubiquitous connectivity. However, that’s all it does well. We are attempting to stretch 1970s technology beyond its design limits.

What happened back then was a deal with the network protocol Devil: we make many expedient, rather than good, engineering choices. A global address space, no abstraction API, control systems moved outside of the network.

These allowed us to quickly grab the statistical multiplexing gain of packet data, and dramatically drop costs. However, as a form of resource management, the Internet is akin to MS-DOS in its sophistication. Past expediency means increasing complexity, constraints and costs now.

What we need is something more modern, a bit like OSX or Windows. These support many concurrent applications, managing the appropriate allocation of shared resources to each. The operating system keeps them secure from both each other and outside attack. The Internet currently lacks this kind of performance and security isolation.

Furthermore, a PC operating system ensures resilient operation of the whole computer even when resources are tight or things go wrong. These are the properties we want of a future Internet that becomes safety-critical.

The journey to a better Internet
To achieve these requires us to undertake a difficult journey, in three distinct, but related, ways.

Firstly, we need to fix a basic science deficit. With cars, we quickly understood the chemistry of internal combustion. With computers, the theory in the 1930s preceded the practice in the 1940s. Until now, packet networks has lacked a robust scientific underpinning.

That is because we have been missing a whole branch of mathematics to model them. That gap is now fixed, but it takes a painful un-learning exercise as much of what is in the textbooks is wrong or misleading.

Secondly, we need a re-architecting of the Internet around sound fundamental principles, reversing the many technical mistakes of the past. The absent security, performance, mobility and resilience capabilities need to be incrementally added back in. We know how to do this, but the several key replacement technologies remain nascent.

Thirdly, we need regulatory policies that reflect that the Internet has not reached some stable end-state, but is very much a work in progress. Anything that ossifies what we have is highly undesirable. What we need are policies that enable many new experiments to take place, and existing players to be replaced by more innovative ones.

There are many vested interests making money from the ways in which the current Internet is broken. Acknowledging that what we have is deficient involves a collective loss of face. Expect much resistance to such change.

No longer just a prototype
What is in prospect is a radical leap in capability and cost, due to new science and better technology. This offers a rare disruptive opportunity. The future Internet can offer both richer economic models and improved technical outcomes.

Indeed, we are at an inflection point. The prototype Internet we have enjoyed until now has followed a “purpose-for-fitness” approach, whereby we took whatever we were offered and discovered what it was useful for. Going forward, the Internet needs to offer “fitness-for-purpose”, whereby it says what it does, and does what it says. This is the essence of what any true infrastructure offers, whether physical or digital.

Then the Internet would no longer be a prototype that escaped from its lab. It would instead become a matter of ordinary engineering. Slightly dull, but absolutely essential to society.

This article was first published as a Future of Communications newsletter. You can sign up here for free.