Ten ways the Internet could be better

I am chairing a panel at the TMRW event in Manchester on 20th May. They have titled the session “The Internet is Broken: Who’s Going to Fix It?”. This has made me think: knowing what we now know, how might the Internet become better in future?

More transparent

Today’s Internet has no “introspection” capability to look inside it. I’d like an Internet that both said what it did, and did what it said. Is this access link fixed or wireless? Which jurisdiction is it associated with? What’s the performance promise (if any) on offer? Is there a captive portal splash screen? Is there a service quality issue right now?

The APIs for these don’t yet exist. I wish they did.

Cheaper for everyone

The essence of broadband is sharing a common resource. I’d like an Internet that intensified the sharing, lowering costs for everyone. Today’s Internet is too reliant on idleness to work, which defeats the purpose of packet networking. It also carries all the bulk traffic with the same cost structure as real-time data. Better scheduling of that data could drive significantly lower costs.

If we simplified the Internet it would be easier to automate and cheaper to operate. The piles of kludges we have built over decades interact in unpredictable ways, throwing up new problems. There is now exploding complexity for network engineers that we need to tame and automate.

More reliable and resilient

It is true that packet data combined with the right transport protocols allows you to recover from “damage”. However, that recovery process is slow and impacts the user experience. A better Internet would allow for local outages to be detected and routed around in milliseconds, not seconds.

Today’s Internet is very poor at what is called “multi-homing”, which allows for several redundant connections to one application. Fixing some architectural issues around routing would make it normal to be able to fail-over connectivity without any impact on the user.

Predictable performance

Who wouldn’t like live voice and streaming video that just worked! It is too much to ask for? Unfortunately, the IETF’s efforts in this direction look doomed to fail.

The Panglossian hope for “too cheap to meter” abundance doesn’t exist. We can’t all have the best performance all the time. We need improved mechanisms and incentives for things to predictably fail in the “right “ (or at least “good enough”) order.

Safer for society

The current Internet is lacking robust intellectual and theoretical foundations. It collapsed completely in the 1980s, chucks up nasty performance surprises, hits painful scaling limits, and suffers for a growing internal “statistical noise”.

A better Internet would be more like the power grid, where we fully understand its operation, can model its safety, and can depend on its sustainable operation. We would be able to automatically isolate any performance problems with total certainty over cause and effect.

Integrated security

The Internet’s design is really a big scaled-up LAN. It has the same basic security architecture as the 1970s research networks it emerged from. Everyone is trusted to connect to everyone else.

Today’s world has a different threat structure to the 1970s, and the costs of failure are higher. I’d like an Internet with proper security controls over association of end points. Denial-of-service attacks should simply not be possible. Intruders would be easily tracked.

Wider choice and no lock-in

When you buy a broadband service today you are locked into a single ISP that provides every broadband service (Internet, VoIP, video, small cells, etc). I’d like to have back the best of the PSTN, where we had a choice of dial-up providers and could switch services. Unlike the PSTN, I’d like to be able to run many of those services concurrently over a single common, shared infrastructure.

Made for mobile

When we went from fixed access to mobile access, we lifted and shifted protocols that were not designed for the purpose. As a result performance when mobile is poor: TCP/IP isn’t suitable for dealing with rapidly-varying radio conditions. We’ve also created a set of hacks to enable mobility (e.g. 3GPP and HLRs) that are disconnected from the underlying architecture.

A radical simplification and cost reduction is possible. A “mobile first” approach to protocol design would have other collateral benefits, such as better battery life and faster hand-overs. These are especially important for sensor networks and wearables.

Extensible by all

Today’s Internet is very successful at creating a “connectivity commons”. Everyone can interconnect without worrying about how anyone else implemented their part of the network. However, this compatibility has serious limits.

As an end user or a local community, I can’t easily add on extra bit of network and truly integrate it into the Internet. I am at the mercy of a service provider, or I have to make my network pretend it’s a single device. There’s little or no common interoperability for service management and operational performance.

A better Internet world allow much more democratic “bottom-up” construction of networks, and make us less dependent on feudal “top-down” providers.

Harmonious and happy

The last thing I’d like is an Internet that engenders more peace and consensus. This is a technical wish for improved technologies of sharing. Today’s Internet has us all polluting a common resource, yet greedily scooping up the best quality for ourselves. I’d also like an Internet that was less conflict-ridden in terms of policy, with a sound scientific basis for technical decisions.

Achieving this harmony requires us to collectively envision a better Internet, and collaboratively move towards it.

How would you like the Internet to be better? I am curious to know.

For the latest fresh thinking on telecommunications, please sign up for the free Geddes newsletter.