Five Ways the Internet is Not Open

Keep the Internet Open” demands the chief Internet propagandist of Google in the New York Times last week. Earlier in the week I saw the same Vint Cerf open the Freedom to Connect conference in Washington DC with the same message. Whilst he rightly opposes the tendency of forces of political authoritarianism and economic rent-seeking to appropriate communications systems, the premise of the article and speech is false.

The Internet is not ‘open’  – unless by ‘open’ you mean “able to support Google’s business model without hindrance”.

The logical fallacy is to conflate the ability of the user at the edge to select devices and services of their choice with the network itself being open. The telephone system, both fixed and mobile, has long allowed us to attached devices of our choice. The arrival of fax and dial-up ISPs showed how the PSTN was open to new applications. ISDN extended that flexibility to sending any data we liked. Yet the phone system is definitely not ‘open’, in the same way an iPad or iPhone is not: you can make choices of applications, but within tightly circumscribed boundaries of a closed hardware platform.

Likewise neither is the Internet truly ‘open’. This is particularly true if we measure it by the practical reality of what we have, rather than the theoretical possibility supported by the protocols in a hypothetical world that does not exist.

Here are five reasons why .

Not open to access

The reality of the ‘edge’ of today’s Internet is that it is locked down and closed. Cellular is always closed and needs payment and provisioning. Truly open Wi-Fi is a rarity that is getting rarer. What this means is that we place high friction in the way of new uses and devices. Unlike power sockets, networks need authentication to access the resource. Your mobile health monitor will stop working the moment you step out of coverage, will never negotiate that splash screen offering ‘free’ access, and the roaming bill will give you a heart attack all on its own.

A truly open system would be more like an information ocean than an information superhighway. It would offer a scavenger class of traffic, at no marginal cost, as a basic ubiquitous societal platform for communications.

Not open to extension

The core routing protocols of the Internet are not open to the users. You are a feudal serf, dependent on an ISP to give you an IP address and a place in the world. You can’t get access to the router-to-router protocols and advertise and add on your own routes and subnets. There is little point in neighbours joining themselves together because the protocols and deployed equipment won’t use those routes. You’re effectively forced to use ‘official’ routes in the ‘official’ controlled and closed hierarchy.

A truly open system would empower users at the edge to extend and modify the system; an “outernet”, not just an “internet”.

Not open to negotiation

When you connect to the Internet, there may be implicit terms (like a sign on the wall), or explicit ones (the Ts&Cs of the ISP on a splash screen). These commercial and technical terms are not machine-readable, and there’s no menu or leeway. You’re going to eat whatever you’re given, even if it’s DPI sprouts boiled with capped cabbage.

A truly open system would have an API to query the terms of connection, and negotiate them. it would allow me to advertise terms and conditions on which I will connect. It would enable collaboration both between the user and network over sharing the quantity and quality available in the network resource, and also between users (so someone who wants more resource can negotiate it with other users).

Not open to diversity

Different types of traffic have different needs – notably, different tolerance for loss and delay. The Internet comes from an environment where the dominant applications are file transfer and terminal emulation. These are relatively undemanding. As we multiplex together voice, gaming, M2M, multi-party video and swarmed P2P downloads the diverse needs begin to overwhelm what an Internet-style architecture can deliver. The Interent is not really ‘open’ as I can’t access the algorithms in the queues upstream of me and ask them to manage my traffic differently for me. FIFO is not natural or neutral.

A truly open system would recognise a diversity of needs and support cooperative means of accessing and sharing the underlying resource. TCP/IP is not the protocol suite to achieve this. Instead we need protocols that allow the edges to collaborate to manage contention: an “interflow” rather than an “internet”.

Not open to dissent

Rather like a Stalinist state planner, the Internet requires a single, simple plan for global routing. You’re either part of the system, or out of it. (There is a Siberian exile called ‘private subnet’.) This neatness appeals to the engineering mindset, but the IETF represents only a narrow segment of humanity and reflects incumbent ideas of how networks can and should work.

If you have a different idea of how things like mobility should be modelled, tough luck – you’re stuck with dependency on a service provider, rather than being supported via co-operative protocols run from the edge in collaboration with the network.

A truly open system would understand the difference between hard constraints imposed by physics and mathematics, and policies that are subject to different ideas. Why can’t I and the other nodes attached to a WiFi access point agree to a different way of sharing the resource, and upload some new algorithms? Why can’t we upload and agree on our own routing protocol and extensions, within a larger framework that protects and isolates parties resources contending for finite resources?

But IPv6 fixes all this!

No, it doesn’t.

Free or open?

What really matters is not whether something is open, but whether it supports freedom. A headline “Keep the Internet Free” would be a better rallying slogan, as it focuses on what really matters. The PSTN is a social compact as much as a network, and has grown laws and regulations to codify a high degree of freedom for its users. Where’s the social compact for the Internet? It’s got lost in the dreamy utopianism of many of its proponents.

However, there is a model in the software domain to draw inspiration from: the GNU General Public License (GPL). This is a sophisticated legal and social agreement to ensure that the pool of ‘free(dom)’ software keeps growing and cannot be enclosed. What would a similar GNU General Internetworking License look like? And how could that encode freedom to connect beyond the dreams of those who fathered and founded the Internet?

[Editor’s note: Martin’s on fire, this is the second article of his we’ve run in last week. This article was originally published as a Future of Communications newsletter. You can subscribe at (and you really should.)]