Replication Is Not Fragmentation


You can listen to or watch this article here:


This article is a response to Vitalik Buterin’s “The Meaning of Decentralization” Essay of February 6th, 2017.

In his essay Vitalik wrote:

“Logically centralized” comment by Vitalik in his “The Meaning of Decentralization” essay.

Above, Vitalik is confusing replication with centralization. Therefore, his conclusion is incorrect. Blockchains are, in fact, politically, architecturally, and logically decentralized.

The fact there is a “commonly” agreed identical state every 15 seconds, and that the aggregate of network nodes, when achieving consensus, behave “like a single computer”, is really swarm behavior, which is a type of emergent behavior, between very different and diverse node operators. There is no actual single state, much less a “single computer”.

“From the perspective of the mathematical modeller, [swarm behaviour] is an emergent behaviour arising from simple rules that are followed by individuals and does not involve any central coordination.” Source: Wikipedia.

Vitalik conflates the terms “common” with “single”, and in doing so he comes to a wrong conclusion of “centralization”.

The term “common” means rules and information are present in many nodes, which is, by definition, decentralization.

The term “single” means a unique point, e.g. one or very few nodes. But, as said above, the execution of code and replication of the database is done in many nodes, in fact, thousands around the world.

When thousands of independent and untrusting nodes in a worldwide network, across multiple jurisdictional and cultural boundaries, voluntarily use the same rules to achieve consensus on an identical state, at exactly the same intervals, that state is actually replicated, which means it is decentralized as it is not physically located or custodied in one or a few nodes.

It is understandable that replication of the same information may give the false impression of centralization, thus induce to bad decision making, because the information is identical in every node. But the reality is that the information is in different physical places, therefore there is no central point nor vulnerability that may put its integrity in any danger. That is actually one of the central tenets of blockchain security.

This conceptual error leads Vitalik to not see the security of replication and confuse it with the insecurity of fragmentation. It is not mentioned in the essay, but this happens when he uses the same flawed logic when promoting sharding in Ethereum as secure.

Fragmentation of the database is, in fact, a tendency to centralization and totally opposes basic security assumptions of blockchains:

Replication is not Fragmentation.

 As can be seen above, if all the information is replicated in all nodes in a network, then one or a few nodes can fail, but the integrity of the information is still preserved as long as there are remaining functional nodes. To the contrary, if the information is fragmented, just by nocking down a few nodes, say, green or red, entire segments may be lost, e.g. green or red would disappear from the system, which destroys the integrity of the information.

Conclusion

The only “logical” conclusion above is that truly secure, public blockchains are logically decentralized precisely because there is one commonly agreed replicated state in many computers around the world. And, the fact the system behaves like a single computer is just an emergent feature of the voluntary and individual desire of node operators. Not centralization.


Code Is Law

Author: Donald McIntyre

Read about me here.