Networks have different structures.
Architecture is Politics (and Politics is Architecture). The structure of a network itself, more than the regulations which govern its use, significantly determines what people can and cannot do.
Mich Kapor, 2006, “Architecture is Politics”
I’ve been thinking about how the structure of a network shapes the flow of money, and the distribution of power within that network. My sense is that different network shapes have different “flavors”, in the sense that they tilt toward certain behaviors.
Centralized networks tilt toward authoritarian
Centralized networks tilt toward centralized governance. The money and power are centralized, so governance is too. Benevolent monarchy at best, totalitarianism at worst.
What is a centralized network today? Most websites and apps can be seen as centralized networks. For example, facebook.com is a central hub that coordinates between many spokes. Suburbs are also centralized networks, with hub big-box stores providing food and fuel for spoke tract homes.
There is a lot of centralized software. It is easier to build. It is more profitable, too.
Aggregation and SaaS give centralized services a defensible business model.
The central hub acts as a single source of truth for coordinating, and allows you to side-step extremely difficult synchronization problems.
You can roll out changes and fixes to everyone at once.
Centralization is often necessary for new product innovation. Innovations often start out vertically integrated in order to line up all the pieces in the stack.
On the other hand, centralized governance confronts the same challenges as any other form of central planning…
One size must fit all. Centralization means everyone shares the same infrastructure. There is limited capacity for pluralism or bespoke local solutions.
Bottlenecked by bureaucracy. We can think of the central bureaucracy as a CPU who’s job is to compute policy decisions for the rest of the network. The complexity of the problems that can be dealt with is limited by the complexity of the central bureaucracy.
Ashby’s law of requisite variety:
If a system is to be stable, the number of states of its control mechanism must be greater than or equal to the number of states in the system being controlled.
Another way to see this is that a centralized bureaucracy does not have requisite variety to govern organic growth. The edge is more expressive than the center, and changes more quickly. This forces the center to limit the variety at the edge in order to maintain control of the system.
Heaven is high and the emperor is far away. When gazing out from the center of the empire, the troubles of the periphery seem abstract. Everyone looks like an ant, from a distance.
Few feedback mechanisms. Those at the center have all the power, yet are far away from the ground truth. Those at the edge are closer to the ground truth, yet have no access to power. Because people at the edge don’t have power, it is difficult for them to create feedback mechanisms to pass information back up to center. When the center puts feedback mechanisms in place, decisions are bottlenecked by the limited CPU capacity of the central bureaucracy, and the mechanisms cannot evolve from the bottom-up to match changing conditions.
Reality debt builds up between center and edge. Cheats, one-off compromises, and corruption are used to deal with especially acute “errors”. As centralized internet services grow to global scale, we see these errors compound and erupt in strange ways.
Natalia Krapiva 👩🏻💻🏠 @natynettleBREAKING: @Apple and @Google just simultaneously pulled Russian opposition voting guide app on election morning. It gets worse. THREAD 1/ Reuters: https://t.co/TBAoeIW6aA https://t.co/n8zITfDGMD
While centralization first evolves for reasons of efficiency, the purpose of the resulting hierarchy soon shifts toward its own perpetuation.
The original purpose of a hierarchy is always to help its originating subsystems do their jobs better. This is something, unfortunately, that both the higher and the lower levels of a greatly articulated hierarchy easily can forget. Therefore, many systems are not meeting our goals because of malfunctioning hierarchies.
Donella Meadows, 2008, “Thinking in Systems”
Once power and money are centralized, there are a number of motivated actors who might want to keep it that way.
It’s easier to extract rent from a centralized system.
It’s easier for governments to regulate a centralized system.
It’s easier to surveil a centralized system.
P2P networks tilt toward individualism
A P2P, or distributed (C) network is often pictured as atomized individual actors who network together in an ad-hoc way, in acts of voluntary exchange. This picture may be construed as anarchist or libertarian, depending.
Philosopher Charles Taylor refers to this kind of structure as a Direct-Access Society. In a Direct-Access Society, people are conceived of as individuals, disembedded from social hierarchies, free economic, or voluntarist agents. These agents interact with each other directly, unmediated by any social hierarchies.
An idealized P2P network tries to prevent any locus of power from forming in the structure of the network. The aim is to distribute power evenly between actors. No one is in control. There is no authority to appeal to.
No one is the boss of a p2p network, so there is no single source of truth. This makes P2P engineering a crash course in philosophies of relativism. Yet, consensus must form around some centrality. In fact, any kind of linearity implies centralization. Ordering/ranking is normative, can be done in many different ways, and thus linearity produces conflicting accounts. These conflicts can be resolved by an authority, or else some “centralized” shared protocol (e.g. a deterministic algorithm). We could see p2p protocols as another kind of hub-and-spoke model, but the centrality is in the protocols, rather than the power accumulated under an authority.
We could say P2P centralizes the system around the protocol, in an attempt to distribute something else (control, resources, or power). This means governance has to be encoded within the protocol itself. P2P protocols may reach for direct democracy, market mechanisms, and other individualistic, or atomized mechanisms to solve problems of governance and consensus.
This can also have downsides. No one is in control means no one is in control, for better and worse. A decentralized system can have collective action problems, where each actor freely pursues their own ends, but the total emergent outcome is bad for all. Bank runs, financial crashes, climate change...
When a subsystem’s goals dominate at the expense of the total system’s goals, the resulting behavior is called suboptimization.
Donella Meadows, 2008, “Thinking in Systems”
Federated networks tilt toward pluralism
Federated, or decentralized (B), networks tilt toward a patchwork of private and public shared spaces. They are a hybrid model, where many small-to-mid-sized hubs agree to cooperate together using shared protocols.
What is a federated network today? Email is a federated network. You can send emails between different providers because they interoperate over shared protocols. Mastodon is a federated network. So is Federated Wiki.
Federations are pluralistic by nature. Different groups can have different governance models, tailored to the goals of the community. This grants federation more variety than centralized one-size-fits-all networks. It also grants federation more variety than p2p networks, since a protocol can never have requisite variety to encompass the complexity of human nature.
Federation also suggests small, shared mutable worlds with clear trust boundaries. This begins to sound like an Ostrum Commons.
Managed by locals
Community makes its own rules
Community can monitor behavior
Graduated sanctions for those who violate community rules
Cheap, accessible conflict resolution
Self-determination of the community
In large commons, multiple levels of management
So, my sense is that federation can be a natural fit for communal-commons modes of cooperating.
At the same time, as a hybrid model, federation shares in the problems of both centralization and p2p. Suboptimization and tragedies-of-the-commons (spam) can happen at the protocol layer. Abuse of power can happen at the fed layer.
It’s more complicated than that
It’s rare, bordering on impossible, that an evolving network is evenly distributed (C). Real networks look more like (B), with varying degrees of centrality. Really we’re talking about about a spectrum, from more to less centralized.
In fact, centralization emerges organically in all mature ecosystems. Why?
Centralization is efficient. Hubs evolve organically within networks to facilitate routing. If you’re trying to get a message from node A to node Z, it’s cheaper to route through a hub with connections to a large number of nodes.
Hierarchies evolve from the lowest level up—from the pieces to the whole, from cell to organ to organism, from individual to team, from actual production to management of production. Early farmers decided to come together and form cities for self-protection and for making trade more efficient. Life started with single-cell bacteria, not with elephants.
Donella Meadows, 2008, “Thinking in Systems”
Centralization has economies of scale. It is often the case that the bigger something gets, the cheaper it is to scale. For example, the web centralized around economies of scale in server farms, so now AWS runs a large chunk of the web.
Centralization snowballs. It’s more useful to connect to a node with more connections, because you can communicate with more nodes. So the more connections a node gets, the more it gets. This is called preferential attachment, and it happens around any kind of more-gets-more Matthew Effect: connections, capital, power. Preferential attachment results in networks with power law distributions.
Federated networks are always in danger of collapsing into centralized networks for this reason. P2P networks aren’t immune either. All networks are under constant evolutionary pressures to find ways to centralize. Centralization tends to happen in unexpected places — usually places outside of what the system designer would consider to be the boundaries of the system.
This makes it difficult to escape centralization.
There are also other natural tensions here. For example, the right kind of centralization can make a network extremely robust against error, even while making it vulnerable to attack.
Real systems are also layered, and different layers evolve at different paces. Punctuated Equilibrium is happening at all of these layers.
Evolved systems end up being tapestries of centralization, decentralization, and federation, for these reasons.
Subconscious as a cozyweb commons?
Given all these tensions, it may be more helpful to view network topologies as a design space, where different models, and mixed models can be designed to produce particular outcomes.
My current conceptualization of multiplayer Subconscious leans toward a network of small cozyweb communities, modeled after an Ostrum’s Commons… kind of like a wiki. Getting there is certainly more complex than choosing a network topology. Yet I do think the topology of the network is relevant. Perhaps federation can play a part, and p2p too?