What is decentralization ? (A)

Info-topological conception

Imagine an engine running an 18th century factory, a single machine driving work-stations throughout. By the common presumption, that would be thought ‘centralized’. But where exactly are the bounds of the engine? The big cylinder in the basement? Or cranks on each desk? If the latter, what then is the centralization? … Or imagine the engine scaled up to the size of the planet, though each piece being no larger, but instead connected by a network. The whole runs as before, and it now seems, by the common view, like something ‘decentralized’. But being centralized or not cannot change just by changing the size, can it?

You have to understand de/centralization in an abstract way: about ‘control’, or ‘influence’. It cannot be about whether there is an ‘organisation’, ie a corporation or government, or set of people in command, just as it cannot be bounded in size and shape like a physical machine. Because such a definition will mean you can easily escape from a label of ‘centralized’ without meaningful change: if centralized is taken to mean one computer, or one office building, or one corporation, then to decentralize it we need only send all the employees to work from home, or have multiple computers, or whatever other superficial alteration, and voila, it is no longer centralized!

Merely running on a ‘decentralized’ platform also does not make something decentralized. Imagine that we identically port Facebook (exemplar of centralization) to an Ethereum-like platform (‘decentralized’ turing-complete), fully automate it, fire the staff, then switch the implementations overnight. Everyone uses Facebook just as before (well, it is a million times slower, but ignore that), and it continues to dominate the internet. But it is ‘decentralized’ now, so everything must be OK, right? … So what went wrong? It seems unsatisfactory, because, as above, it is a result of too material an idea of what de/centralization is. ‘Decentralization’ cannot have only an internal, technical meaning – the distribution over computers, the parallelising, something algorithmic. The true concern is instead external, the interaction between a system and its participants or users: what affects what, what is the source of information, and thus control.

So we must think of de/centralization more abstractly, as about the shape of the flow of control, the info-topology ‒ like the branches of an invisible tree that forms an info driven causal network. Centralization is the expression of shared information, in coherence, pattern, organisation. The minimal case is a single info source that disseminates that info to two or more other agents ‒ there is a centralization: an asymmetrical dependency in the relation of behaviour of a set of agents. Centralizations can be large or small and they may vary in the detail and occasion of control exerted,

‒ ‘But flocks/species/language has no center/leader.’
They have an underlying pre-set informational centralization ‒ in physics, genetics, culture.
‒ ‘But those systems evolve, unfold.’
They drift, unpredictably ‒ but that is disorder, not systematic.
‒ ‘No, it is structured …’
Then we are back to some underlying centralization again. Structure is produced by causal focuses/concentrations. Pattern/order/coordination of multiple elements can only flow from single seed-points.

Wherever people/agents/elements act coherently, all following some pattern, there is a centralization – no matter how abstract it might appear, the centralization is the single information source that determines the pattern they all follow. Decentralization is just the absence of that: where agents lack a coherent pattern, or the form and degree to which they do.

Virtual/fixed centralization

From that, one can see that ‘decentralized’ stuff as commonly thought ‒ bitcoin, etc ‒ does not ‘remove centralized control’, it just freezes it at the point in the past when the algorithm was set, since that is the info source everyone is following.

Think of these scenes 4-dimensionally ‒ as in time as well as space ‒ and imagine the causal chain/network propagating over events through the volume. Decentralized stuff does have a ‘single point of failure’, but one located at a single earlier point in time, and so inaccessibly locked in the past. The ‘point’ of inception of the algorithm is when the behavioural structure was defined, but since the custom is to promise this as immutable, it will then not change. And what is called ‘centralized’ is distinct from that by having that point extending along a line in spacetime ‒ carried through time and continuing its influence, and so continually accessible. If you set up an organisation to maintain or enforce a behavioural structure, its ongoing capacity to intervene is also itself open to influence. What is called a ‘single point of failure’ is really a line in the spacetime causal volume.

You could call the common idea of ‘decentralization’ a ‘virtual centralization’. The centralization is in an algorithm, instead of a conventional material form like an organisation of people. And that rarefied sense makes people think it is not really there, or that it is somehow special or magical. In truth, this ‘decentralization’ just takes what it claims to object to, and substitutes it with a newer more abstract kind of machinery, that does the same/similar thing.

Have a look at some interviews from https://redecentralize.org/interviews/ , and it is easy to see the centralizations hiding in plain sight:
‒ “[Someone] talks about [Something], a real-time framework for making decentralized apps.”
‘Framework’: there is the centralization.
‒ “[Someone], founder of Protocol Labs, describes the [Something] File System, a decentralized content distribution protocol.”
‘Protocol’: there is the centralization.
‒ “[Someone] describes decentralized marketplace [Something].”
‘Marketplace’: there is the centralization.
‒ “[Someone], founder of [Something], a platform for making ‘Internet of People’ apps.”
‘Platform’: there is the centralization.
And so on … All these proffer a centralization. They want multiple users for their framework/protocol/platform/marketplace, do they not? So there is the centralization: that is the single structure multiple people are following.

Now compare the two: pre-set/frozen-in-the-past-centralizations versus active/ongoing-centralizations. Given its inaccessibility and immutability, ‘decentralized’ means non-adaptable, and that means ‘decentralized’ is a strictly less capable structure. Cryptoledgers, as a combination of algorithm computers and humans, form a self-sustaining system ‒ a pattern of activity that endures over time. But it is not adaptive. It is only a fixed pattern, unresponsive to its environment. But ‘centralized’, with its ongoing means of influence, must also have the ongoing potential of being influenced, and of being adaptable. It is thus more sophisticated, since the abstraction is expressed and carried over the more moving pieces of a dynamic, changing, environment/context. So, contrary to the publicity, ‘centralization’ is the more advanced and capable, and ‘decentralization’ is the regressive.

Related

• "What is decentralization ? (B)"; hxa7241; 2021 / 3056 words / article.
• "Decentralization's deficiencies"; hxa7241; 2017 / 888 words / note.
• "The illusion of decentralisation"; hxa7241; 2015 / 1048 words / article.