politikblockchainWhat it means for a network to be decentralised
"Decentralised" has become a label that either sticks or does not, as though it were a switch. A network is or it is not, and the conversation ends there.
It does not work like that. It is a measure, not a state, and it is not measured once: it is measured layer by layer. A network can have consensus spread across thousands of participants and depend on a single software provider. Another can have the software spread out and decision-making power concentrated in five wallets.
Not a property, a distribution
The useful question is not "is it decentralised?" but "what would it take for someone to impose their will?".
Put that way, the answer is always a number. How many participants would have to coordinate to censor a transaction, to reverse a block, to change the rules. That number is what matters, and it is usually far lower than the rhetoric suggests.
The layers where it is decided
Consensus. Who validates, and with what weight. This is where almost every network shows off, and where concentration hides best: thousands of validators can be delegated to a handful of operators.
Software. How many independent client implementations exist and how much each is used. If 90% of nodes run the same program, a bug in that program is a bug in the whole network — however many nodes there are.
Infrastructure. Where those nodes actually run. A network spread across ten thousand machines hosted by three cloud providers is not as spread out as it looks.
Governance. Who decides what changes. It is the least audited layer and the one that most determines where everything else goes.
All four can be measured separately, and the weakest sets the ceiling. Exemplary consensus counts for little if a single foundation decides which version ships.
How it gets quantified
The most widespread way to put a number on it is the Nakamoto coefficient: the minimum number of independent entities that would have to agree in order to compromise a given layer.
It is an imperfect tool — it depends entirely on how you define "independent entity", and a great deal of creative accounting slips in there — but it has a virtue no adjective has: it forces a figure. And a figure can be argued with, compared, and tracked over time.
When someone claims their network is decentralised, the question that organises the conversation is what coefficient it has, on which layer, and who calculated it.
Why this is not only a technical debate
A system that distributes control also distributes responsibility, and with it the capacity to answer to anyone. If three entities can censor transactions, there is a point of leverage: a regulator, a court or a government has someone to address.
That is neither necessarily bad nor necessarily good. But it is a political fact, and it is settled in the same technical layers listed above.
Which is why it is worth looking at the numbers before the declarations. A network's architecture is not an opinion about how power should be distributed: it is the distribution already made.