How is network neutrality like pigeons playing chess?

The idea of “network neutrality” has been in the news a lot recently. Rather than address it directly, I want to offer you some thinking tools to help position the whole debate in the right place in your head.

I have read Nassim Taleb’s brilliant Antifragile, together with other writing of his. He uses three simple terms to describe the constraints that exist in the world: cosmic, ludic and ecological. Let’s take an example of the stock market to explore them.

The three constraint on everything, everywhere

The first are cosmic constraints. These are those imposed by physics (or God, depending on your outlook on life and the universe). The obvious one is the speed of light, which isn’t up for negotiation.

In the case of the stock market, high-frequency trading is constrained by latency. It’s why traders are willing to spend $300m for specialised links to shave a few milliseconds off the time between New York and Chicago. However, they can’t keep on endlessly shaving that time down; there’s a lower cosmic constraint (of around 4-5ms).

The second are ludic constraints. The word comes from the Latin ludere, which is to play (a game or sport). Games have rules, which impose constraints. These are typically mathematical in nature. A game of chess, for example, has a state space and set of legitimate moves which can be described using a mathematical model.

The stock market is a kind of game, with very simple rules. Buyers and sellers place orders, and the market matches them up. The quantity of stocks sold minus those bought always equals zero. You can’t go to the market-maker, complain about your financial losses, and have them create new stocks for you. That’s not part of the game.

Finally, there are ecological constraints. These could be ones of technology, economics, policy, etc. They may be absolute, in the sense we cannot control them. An example would be human nature. They could equally be pliable ones, such as how we make policies about how humans are allowed to behave. Or they might be moving targets, like the evolving state of the art in a particular technology.

For the stock market, we have policies against front-running trades and insider trading. We charge fees for using the market, and the amount of the fee is constrained by the value the participants get from using the trading platform, as well as the cost of running the IT systems.

These constraints form a strict hierarchy: the cosmic ones come first, then the ludic, after which come the ecological. There’s no point making policies about the stock market that require the rules of physics or mathematics to be broken. As such, these constraints filter what you might want, and their intersection defines what you can actually get.

The constraints have a hierarchy & act as a filter

In our world of broadband, the cosmic is quite simple. For example, Shannon’s limit imposes a hard constraint on how much information we can send over a channel. Network neutrality exists in the ecological world of policy-making. Analyses of network neutrality issues assess the likely impact of different policies on markets and users.

That leaves us with the ludic, which is where we get into deep trouble. Broadband is, by definition, packet-based statistical multiplexing. Its ludic nature is given away by its name: it is like a statistical ‘game of chance’. That means the policies need to reflect the constraints on what is possible imposed by this statistical multiplexing process.

The problem with the network neutrality debate is that it is being conducted in a bubble that is often disconnected from how that game of chance works, and the actual ludic constraints that exist. There are assumptions being made about the sustainability of the current broadband model’s performance and cost, or that of alternatives.

These assumptions are not backed up by hard science. The number of lawyers with suitable higher degrees in stochastics is (probably) zero. Indeed, every single policy paper and book on network neutrality that I have ever read has a flawed understanding of the nature of the resource being debated and its trading space. In particular, they fail to capture that the constraints are shifting from ecological ones (‘speed’ dictated by technology) to ludic ones (ability to schedule and distribute contention).

As a network performance professional, engaging in this “debate” feels rather like playing chess against a pigeon, who “knocks the pieces over, craps on the board, and flies back to its flock to claim victory.” The basis on which people justify their implicit ludic views is so flimsy it is “not even wrong”. As the link says: “The phrase implies that not only is someone not making a valid point in a discussion, but they don’t even understand the nature of the discussion itself, or the things that need to be understood in order to participate.”

That calls for a different kind of debate, where the mathematicians get a say at the appropriate point. Their job is to identify and communicate what these ludic limits actually are. This is a non-negotiable precondition to rational participation in “downstream” discussions about the ecological issues.

Using this knowledge, the players who went to law school can then interpret their common carriage principles into feasible ecological policies. It is not the job of the mathematicians to design the market structures, or set the political agenda. But to continue making policy in ignorance of the ludic limits of broadband is a crappy bird-brained idea.

To keep up to date with the latest fresh thinking on telecommunication, please sign up for the Geddes newsletter


  1. A better analogy would be a child, not a pigeon, knocking over the chess pieces.

    Because you’ve overlooked what happens next.

    “Let’s play a different game”.

    And – if/when what you say about the ludic realities of Neutral broadband is correct – there’s another game to play here too. Ultimately, if the mathematics don’t support the type of multiplexed broadband that politicians (and by extension the voting population) wants, what I think is most likely is a shift to *non-multiplexed* broadband.

    For example, the “rules of the game” would be totally different, if all distributed networks ran on end-to-end optical (or tight-beam microwave) circuits. Now in engineering terms, that’s difficult, expensive, impractical… but as I’m sure you’d agree, those are ecological problems and therefore secondary to the ludic ones.

    The interesting thing is that this shift back from multiplexing towards circuits is already occurring in patches. 60% of cellular base-stations are backhauled with microwave, point-to-point. You talk about the very tight scheduling requirements of small-cell backhaul being a key drivers of better contention-management. Yet some vendors now talk about using >20GHz microwave instead, using reflections and diffraction if needed to overcome non-line-of-sight problems.

    Don’t forget, multiplexing comes in lots of varieties: time-division is but one. Most of the world’s cellular networks use frequency-division multiplexing to separate up-and-downlink flows, while beam-forming uses spatial multiplexing. Air and glass have orders-of-magnitude different economics to copper – and can also support multiple independent flows in the same physical space, notwithstanding a bit of interference & maybe quantum tomfoolery.

    This is where I think Taleb gets it wrong. I’ve read Antifragile too. And I think what he & you both miss is that if the Ludic rules of the game represent hard limits, then perhaps there is the option to play a new game. You may even find that regulation *forces* a new game to be played. In finance, high-frequency trading is being regulated because it “causes volatility”. Public drinking in bars causes “volatility” too – yet Prohibition just changed the game to speakeasies, with an anti-fragile side-effect called “jazz”.

    Maybe a you should think of Net Neutrality as just a light and perhaps unintentional form of Prohibition on packet-based statistical *time-division* multiplexing. Maybe it *is* knocking over the pieces on the chess-board. But given the smarts in the network and computing industry, we’ve already started setting up the Scrabble board (I nearly wrote Monopoly) using microwaves and photonics. Mathematics might determine the rules of a game. But Physics allows new games to be invented.

  2. Martin Geddes says

    There isn’t such a thing as “non-multiplexed broadband”, so the rest is a non-sequitur. What you describe are a variety of ecological situations all constrained by the same fundamental ludic limits. Even switched lightpaths and beam-formed wireless! QED.

    Any “new games” are all still subject to the same stochastic rules: unless you are God and can change the constraints of physics and maths, there’s no escape.

  3. Dean, Martin – great debate and I do love watching a ‘battle of the titans’ unfold 😉

    I would say the question here is really one of scarcity/abundance, which are ecological constraints (right?). If high-quality connectivity was abundant (low loss and latency achievable for, say, free) – the ‘Bubley switched-lightpath world’ – then this would lead to a different set of (again, environmental) policy decisions than a scarcity scenario, where a $30 broadband ARPU is not sufficient to satisfy all demand, all the time, at the desired quality.

    It’s like the basic rules of the game are the same, but the pieces and cards have different values and the dice is a different shape (so the house rules have been updated accordingly).

    So I’m arguing myself into agreeing with Martin here – I think. But Dean could be right that new approaches could change the ecological conditions so radically as to make the ‘stochastic rules’ less relevant to the overall outcome.

    But we’re getting 10 years ahead of ourselves here, I guess…


  4. Martin Geddes says

    Richard – thanks for chiming in. The existence of scarcity or abundance is the result of the composition of the cosmic, ludic and ecological constraints. There are “outermost” constraints, such as the maths of finite queues, that don’t change. Then there are “innermost” constraints that we may be able to vary (e.g. QoS algorithms available).

    The long-term dominance of ludic constraints is a simple by-product of the “success” of ecological growh of transmission technology capability. As capacity becomes abundant, it has the side-effects of decreasing the isolation of rivalrous flows being multiplexed together. Thus resource schedulability becomes the limit, not capacity. This is described in more detail in the article “The third epoch of telecoms”, which is here:

    Some of the changes Dean describes would, in fact, accelerate this process! We could have even less isolation between users and flows. For example, in 2G, it would take a very large number of handsets to saturate a single cell tower’s backhaul, whereas in 5G a single handset may be able to saturate multiple towers at once.