I have written about this myth many times before, both as posts and comments, but seems like people prefer believing in myths instead of understanding reality. Decentralisation is the biggest myth in crypto. I have seen far too many articles start with,"decentralisation is hard to define but..." followed by some miracle related to the "power of the people". What is totally absent is the reality that the power only belongs to some people. What is worse, is that the algorithms that allegedly define "decentralisation" are perfect for being hijacked by agents of collusion. Just as democracies can easily become tyrannies, so decentralisation leads to centralisation - is just a matter of time and will.

What Does Decentralisation Mean?

Firstly, let us inject some clarity into what "decentralisation" actually means, what it should mean, and how the concept is totally ripe to being abused. "Decentralisation" is not a noun; it may look like one, but this is part of the camouflage. "Decentralisation" is an adjective, or, more precisely, an adjectival noun. It is totally meaningless without reference to what is being decentralised. This is a deep source of confusion, as I rarely see this qualifier. This leads to people arguing over the definition of something that can be many things - and that is a total waste of energy.

There are two main, and separate, uses of the word: one is decentralised governance; the other is decentralised network. These are very different things. Sadly, the construct known as DPOS (Delegated Proof of Stake) merges the two into one, and hence encourages the confusion. Proof of Stake (POS) and Proof of Work (POW) systems also have varying levels of cross-contamination between the two forms of decentralisation.

Before going forward, let me pause to add that the idea of decentralisation can be used all the way down the levels of a network, and all the way up to the social level - from metal to meatware. If one tries this exercise, listing all the network components and systems and functions, it becomes obvious that "decentralisation" is a terrible concept. Hence, I've written before about changing one's perspective to discuss such sub-systems in terms of dependencies - that is a far more useful concept to manage.

For example, a website requires a domain name; hence, it is dependent on the domain naming authority that manages the top level domain, be it dot-com, dot-blog, dot-io etc. We have even seen one website that many of you have used lose its domain name because it was registered by a person who had left the project. That's a dependency! Can that dependency be reduced, or even eliminated? That is a serious admin question.

Another example, imagine you have a distributed network of nodes acting as Graphene blockchain witnesses. That appears to be decentralised. But look again, they are all using the same server hosting facility (that's a dependency), look closer and they are all on the same physical machine (a further dependency), then you notice they are all run by one human using multiple usernames (that is now a very serious dependency!) At each level of analysis, the dependencies increase while the decentralisation decreases.

This now gives us a good idea of how to define "decentralisation": it is the minimising of dependencies across the whole system.

This gives us some clarity of thought in order to analyse the dependencies and then categorise them as either critical or non-critical. The critical dependencies are those that cannot (as yet) be minimised, while the non-critical ones can be further analysed into ways to make them more decentralised, and hence minimise the dependency on them.

Security and Trust

So let us now get back to what most people seem to be thinking about when they trot out the D-word: governance and/or security. Let us also now focus on DPOS, otherwise every statement will end up with lots of branching other statements.

The security of the blockchain is in the hands of so-called witnesses - other systems may call them validators or miners. They manage the hardware and software that allows them to produce blocks, verify them as valid, and then add them to the chain. As in the example given above, this system can have varying degrees of decentralisation. The ideal is that all witnesses are independent, both at the level of machines and the humans behind them. But how are these witnesses selected?

Before answering that question, let me complete the picture of how a blockchain is run. In this case, let us still focus on the social media chains such as Blurt, and other Graphene chains. Such a chain can easily exist, managed by the witnesses, and yet be doing absolutely nothing! How does anybody else have access to such a chain? Well, if there is documentation you can read that and learn how to interact using a CLI (command line interface). You certainly won't see any images or videos, and finding posts will be a fun challenge.

On such a chain, the witnesses can play with each other; this is the sort of setup when Steem first started as a coin mining protocol - the witnesses were miners, just as for Bitcoin. What is needed to enable other users is an RPC node that allows communication to and from the chain. This then also requires some front-end website, or app, that converts clicks into operations that then produce transactions.

We immediately see that we have introduced two new dependencies. Strangely, the vitally important public RPC nodes are not rewarded for their work, unlike witness nodes; thus a crucial dependency has been created that is wholly centralised! I don't wish this to get very technical, but witness nodes could also act as public RPC nodes just by adding the necessary plugins. But enforcing this would require code changes, and would also make the workload much higher for the node operators with potential consequences on their core role of producing blocks.

Also, as we know, anybody can run a front-end so that users can interact using a simple interface. Again, such sites are centrally controlled by whoever wishes to run one. All of this is to stress that in order to run a functioning public blockchain, with which users can interact, requires extra software that is by necessity centrally run. Can either RPC nodes or front-ends affect the blockchain itself? Yes, they can if run by bad actors. They can change the user perception of the chain by filtering content and user actions. We have seen this too on Steem after the takeover. Hence, users need some confidence that the people running such services close to the chain can be trusted. This may sound like a far cry from "trustless protocols" but is the reality of the services that interact with the chain.

Governance

So let us get back to how witnesses are selected. In a DPOS system, witnesses are voted on by users. Each vote is weighted in proportion to a user's vested stake. This is very much like corporations where shareholders vote in proportion to their shares. Just as corporations can have majority shareholders so too can a DPOS system. Anybody can buy enough coins to have a significant voice on the selection of witnesses. Anybody with enough money can even buy a majority stake in a chain. So I ask: in what way is this a decentralised system? It isn't. It may even look like it at the start, but the influence of stake is an obvious incentive to accumulate and control.

And this is where the mixture of network and governance becomes the real Achilles heel of DPOS. Having control over the witnesses is not just about securing the network but is also fundamental to how the chain behaves. The only way to change the rules of the chain is through a hardfork (HF), and such a hardfork must be implemented by the witnesses. Each witness must update their code to match the new rules set out in a HF. If the top witnesses were to collectively refuse to change their code, then any new HF would not be implemented live on chain.

Thus, control the witnesses and you control the very rules of the chain. This is a powerful incentive to accumulate stake, control the top witnesses and thus change the very nature of how the chain operates. Indeed, this can also be used by agents of one chain to destroy a chain run by a competitor.

For some chains, this may require billion of dollars, whereas for some smaller chains this can be done with under one million - maybe even less. So, again, where has the "decentralisation" gone? The obsession with "being decentralised" is just a weakness that encourages collusion among those members who seek to control such a chain.

Steem and Hive are effectively governed by a single entity, with or without varying degrees of collusion; something few people may be aware of is that Steem, before the takeover, suffered from a stalemate of control. This is a long story, but the founding corporation had lost control to a group of other whales. Stalemates can sometimes be good, so no damage can be done, but such a stalemate means that developing the chain can be halted unless done in the interests of a small group. We saw some awful decisions, especially HF21 on Steem, that were blatantly designed to help the whales accumulate even more stake, at the expense of the wider userbase.

I am not privy to what Steem could have become had the developers had full control of the chain rules. However, Dan left to found EOS, which at its core was very similar to Steem. Why was he unable to make such changes within Steem itself? The point here is that no chain starts off with 100% perfect code at genesis, hence every chain needs a way to develop in the future. Taking away that right to develop one's own chain, in the name of a vague idea such as "decentralisation", is the path to ruin. The obsession with "decentralisation" means that even great ideas can be hijacked by any individual or group with enough resources to gain control.

It is so easy to turn decentralised governance into a controlling stake, that the question must be asked as to why should the stake be more important than the founders? Why is it such a great idea that chains can be hijacked for personal profits? Why is it such a great idea that chains should leave themselves open to be hijacked and crashed? And such crashes are not exploit hacks, not code bugs, they are a feature; a feature that seems to have been promoted precisely to allow wealthy individuals to control the whole of crypto space!

Again, think of the weakness of national democracies. Are there any left that are truly the "voice of the people"? They have all been hijacked by huge vested interests that want to control all populations. Think about how the idea of democracy has polluted the world. Read the Confessions of an Economic Hitman to learn how grants, loans and bribes, plus a pledge to be "democratic" are used to subjugate whole nations. "Decentralisation" is the same con. I wish someone would write "Confessions of a Crypto Hitman". Anybody with eyes can see how it is done, but it takes some foresight to see it coming - before it strikes.

The Problem of Collusion

The deep problem with decentralised governance is that there are no existing algorithms that can protect this idea - no way to ensure decentralisation and avoid collusion. This is a hard problem, because the network of colluders exists outside of the network of a blockchain. One can track behaviours, and thereby construct such a network of aligned interests, but it is not easy, and such hijackers are smart. Try to recall the surprise move by Justin Sun to wrench back control of the Steem chain: using an exchange stake to vote on witnesses - something exchangers had promised never to do. It's all about the numbers, so that whatever can happen will happen given aligned incentives.

I have written in the past about using the word "decentralising" as a concept for blockchains. If a so-called decentralised governance chain becomes centralised, are there any protocols that force the rules to re-decentralise? Is there any protection whatsoever from gangs of rampaging hijackers targeting chain after chain? I don't see it. If one can design such rules that maintain a decentralised governance even against centralising attacks then that would be gold. However, it must be reactive enough to avoid the colluders from taking over governance to then change the rules!

Now, that kind of decentralising algorithm may be impossible, and if it is impossible then one is left with only two options: to accept that decentralised governance can lead to a governance coup; or just admit that there is some special stake designed to avoid such a coup. This doesn't alter the decentralised nature of the network, just the governance. This is no different to some corporations that have special classes of shares as protection for founders - those special shares may be sold in the future or even expire. The purpose is to stop a new entity from being hijacked, either at birth or as a growing teenager. We would not be wasting so much energy and attention on such issues if they were crystal clear.

What Do Users Really Want?

I think most active users on Blurt just want a social system that is relaxed, easy to use, simple to earn some crypto and fair to each other. This is, after all, a social network, not some Defi site that is totally finance-orientated. To build a crypto system that is both social and financial has proved difficult - just look around. New sites look great - until they don't! The algorithms that define the financial interactions are hard to encode in ways that are perceived to be fair. Much time and resources have been spent defending the Blurt chain from multiple attacks, resources that could have been spent developing the chain. Users all have the freedom to join, and to leave - nobody is being forced to stay here. But also, nobody should be allowed to hijack the chain for their own ends - and waving the flag of "decentralisation" is the path to giving such people an open door. Those who insist on not understanding this need to look at the propagandists.

So, finally, if we go back to our definition of "decentralisation" as the process of minimising dependencies, does chain governance count as critical or not? Does chain development count as critical or not? On Blurt, the so-called regent-account was created to protect Blurt from precisely what happened on Steem! That account has no financial power and cannot vote on posts. Its only function was to ensure a trusted set of witnesses so that Blurt would not be hijacked. How many users would like to see the downvote reintroduced? How many users would like to see the reward curve reset to the one used on Steem, that hugely inflates whale votes? These and many other changes are easy to do - once one has control of the witnesses. Like I said, the system incentivises and encourages such hijacking precisely because such coups can change the rules immediately. Is that what everyone wants?