Is my crypto decentralized?

[CLains]

When BitShares scales to Visa proportions it will cost something like 5000$ a month to run one Delegate, that's already $500 000 a month cost for decentralization. It goes without saying that you can only increase decentralization so much before these costs will prevent you from being competitive. If we go from 100 to 1000 Delegates at that point in time cost will increase from over 5 million dollars a year to over 50 million dollars. Perhaps it will be worth it, perhaps not - what transaction-fees would you tolerate for this increase in decentralization?

There is no either/or situation here, just economic rationality and navigation between brute realities.

[EvilDave]

Two posts and there could be 50+ Nxters in here who definately wouldn't say it is decentralised.  They would describe it as distributed, in the same camp as Ripple.

But what would be the point of that?


Bitshares definition 'sufficiently distributed' = 'distributed' in mainstream crypto. The core dev refers to it as distributed.

Node owners like StanLarimer seem to be promoting it as decentralized as they see marketing value in it, which feeds into their own bottom line. ~$2500 a month roughly in Stans case.

Cutting to the chase.
The question is: Is Bitshares sufficiently decentralised/distributed to be regarded as a peer-to-peer crypto-currency, or is it in fact a centralised financial system like Ripple or the US treasury ?

101 delegates is not very many on a global scale, these people would become insanely wealthy if BTS achieved global adoption.
I believe that BTS cannot be described as decentralised, simply due to the amount of control of the BTS network being concentrated in so few nodes, and the barriers that have to be overcome to become a node operator.

In NXT, anyone with a fistful of NXT and a spare PC (or VPS) can run a node, forge, contribute to the network and earn fees. Anyone. Obviously, the guys with more NXT will earn more, but there are no barriers whatsoever to setting up and running a NXT node.
Similarly, anyone with a mining rig could (at one time) contribute to the BTC network and mine BTC while securing the blockchain, again no barriers at all, no need to ask permission from anyone.

It's this level of control that makes BTS centralised, they are a (semi-) corporate project, and not a genuine peer-to-peer crypto-currency. Sorry......

[Daedelus]

Here is the kicker, DE  was right when he said this is a solution looking for a problem.

There are already platforms where each node could be run for $30 a year or less, when running at visa levels of transactions.

This capped node system forces a compromise out of ignorance or a desire for profit. It sounds like you are happy to have the economy of 3 billion people in the hands of 101 nodes. Do you run a node?

[CLains]

There are already platforms where each node could be run for $30 a year or less, when running at visa levels of transactions.

How can you avoid this?

At 2000 TPS validation costs equals the cost of running server which include 100 MB/sec network connection, a server with 256 GB of ram or more, high availability (redundant backups), DDOS protection, etc.   These kinds of systems can easily cost $5000 / month or more to operate at scale.

[Daedelus]

By only sending transactions to the next node that forges rather than all nodes. This means netwotk traffic drops to small fraction of what it would be, keeping forging within reach or ordinary people.

[matt608]

No one is calling Bitshares centralised. They are saying it isn't decentralised.

It is this kind of soft headed thinking and misdirection that I have come to expect.

There are 101 nodes in BitShares who can be voted out by stakeholders and thousands of standby nodes at the ready.  That is the fact of the matter.

How many nodes are required to fit for your definition of decentralised?  

The idea that BitShares needs to be able to handle 3 billion users is a straw man.  That level of crypto use could be 20 or 30 years away, no one I know of has claimed BitShares or any crypto can scale scale straight to that level.

Cutting to the chase.
The question is: Is Bitshares sufficiently decentralised/distributed to be regarded as a peer-to-peer crypto-currency, or is it in fact a centralised financial system like Ripple or the US treasury ?

How does anyone in the world create a node and get elected into a network for the US treasury?

[Daedelus]

Decentralisation is a function of the participants in the network. Limiting the number of participants, limits the decentralisation.  I wouldn't call anything that prevents you participating as decentralised.

Don't you want Bitshares to be a global technology? If you do, 3 billion people under 101 nodes is the reality you have to consider.

[matt608]

Decentralisation is a function of the participants in the network. Limiting the number of participants, limits the decentralisation.  I wouldn't call anything that prevents you participating as decentralised.

Don't you want Bitshares to be a global technology? If you do, 3 billion people under 101 nodes is the reality you have to consider.

There is an imposed limit on the level of decentralisation of BitShares, there's no doubts there. 

Anyone can participate though, the standby nodes aren't just waiting, they compete and campaign to get voted in. 

[Daedelus]

Anyone can't participate in security. Only 101.

Campaigning, competing, waiting, on standby don't do anything for security.

[StanLarimer]

Perhaps it will help we agree on a definition of decentralization in this context.

• Some might say something is decentralized if it does not have a central point of control.
• Others might say something is not decentralized until is can not be decentralized any further.
• Many might argue that decentralization is a process that operates on a continuum between the two.

On the prior thread where all this has been debated before, the OPer wrote:

...
With regard to "diminishing returns" and "no real benefit" of increasing above 101 node current cap, most crypto enthusiasts would call this increasing decentralisation.
...
If there is friction in the growth of nodes in relation to the growth of the network, this leads to increasing centralisation.
...

From this, I assume he would agree that there are, in fact, degrees of decentralization.   

So, starting from that point of profound agreement, it may help to propose some new terms to clarify our discourse.  To that end, and without further preamble, I offer the world:

Stan's Super Seven Degrees of Decentralization

1.  Barely Decentralized. A system moves from one point of control to several.  The separation of powers of the US government into three branches could be called "barely decentralized."

2.  Fully Decentralized.  A system is fully decentralized when you can't think of a way to decentralize it any further.

3.  Manageably Decentralized.  A system is engineered to achieve a practical span of control by delegating it downward until people can keep track of everything that is going on.

4.  Plausibly Decentralized.  Decentralized enough to satisfy the general public that control cannot be seized by somebody bad.  Nuclear missile launch authority is deemed plausibly decentralized if two officers are needed to launch having received a verification code from a command authority.

5.  Sufficiently Decentralized.  A system is sufficiently decentralized if it achieves the goals of its designer.  A crypto system is sufficiently decentralized if, in combination with all other design features, it is not possible to gain control of the system to the point of causing it to shut down or malfunction in a sustained and unrecoverable way.

6.  Over Decentralized.  A system is over decentralized if costs more or takes longer to implement than one that is Sufficiently Decentralized.

7.  Under Decentralized.  A system is under decentralized if it is less Over Decentralized than the system you prefer.

A system decentralized to any one of these degrees can be called "decentralized".  If you want to be more precise you can throw in one of the above adjectives.

The BitShares Community,
after a year of iterative design, analysis, redesign, testing and intense public debate
 has concluded that our design is Sufficiently Decentralized
 to trust it with our lives, our fortunes, and our sacred honor.

[Daedelus]

You are the only one left arguing it is still decentralised stan.

As long as there is a cap on participants,  I don't think it will matter how you try to spin it.

[EvilDave]

Fuck, I'm starting to like Stan.  

I'd consider BTS to be somewhere around plausibly decentralised, with good ol' NXT (or, ignoring the mining pool issues, BTC, LTC etc) as being fully decentralised.

While I'm here....could we decentralise NXT any further?

[Daedelus]

He is king of the dodge, sidestep and misdirection 

[toast]

It is crazy to me how people focus on the block producer cap instead of the underlying consensus metric. If you remove the cap and let people produce proportionally to their approval, you would see a higher concentration of blocks produced by a smaller number of nodes.

Having 101 validators is FORCING the network to be MORE decentralized than the other consensus protocols, where 99% of blocks are produced by far fewer than 100 validators. As a stakeholder, your control over block producers is exactly proportional to your stake whereas in other systems it becomes unprofitable to validate unless you control huge amounts of the stake (or hash power in POW, or unique nodes on ripple UNL)

What is the goal of decentralization? It is to make it difficult for anyone but majority stake-vote to decide what the rules of ownership are. To protect the stakeholders from everyone else. The goal is not to make everyone feel like a part of the validation process and pay them for make-work.

[StanLarimer]

When the BitShares devs get a little uppity, I love to remind them that what they are doing is not exactly rocket science.  

But, while trying to find a way to communicate the architecture of BitShares to folks on other forums, I've stumbled on the following description, drawing on my past experience with continuously reconfiguring fault-tolerant flight control systems.  (Yes I wrote a technical report with that title back before there was an Internet or even a word processor.)  See if you buy this way of describing the BitShares architecture:

We tend to get three different attributes mixed up causing endless confusion even among men of good will.

Throughput Scalability and Fault Tolerance and Decentralized Control
are three different concepts.  

Fault Tolerance. We are saying 101 highly-reliable, tested and proven, hand-picked parts dispersed across the globe and selected by the entire owner population is sufficient redundancy to achieve reliable fault tolerance.  The only thing those parts can do is -- do their job to spec.  We can observe their performance and swap them out in ten seconds if they don't perform to spec. So, really, they are just interchangeable slave machines.  Producing the blocks is a mindless task.

Selecting which parts make up the machine is where the power lies.

Decentralized Control.  The total decentralized population of the all owners participate in selecting the most reliable machines to run the network. Those 101 parts have no power over the owners. 101 dispersed redundant parts is a decentralization red herring! That's not where control lies. Those 101 chosen nodes can be completely reconfigured or replaced by the fully decentralized participating owners in 10 seconds.  

We have decentralized p2p control
of a distributed, fault-tolerant computer
implementing an autonomous unmanned company
running a decentralized crypto currency exchange
which produces stable crypto-currency products.

Throughput Scalability is also entirely different.  Any of the 101 nodes can scale up by adding parallel machines, side chains, and a thousand inventions we haven't dreamed of.   When we get to a billion owners and need a thousand machines per node, we can do that.  It will still be decentralized enough to ensure that the 101 (now bigger) parts that make up the distributed, fault-tolerant machine will perform their mindless slave jobs reliability - from positions scattered across 24 time zones.

Those million decentralized owners do not want to have to think about managing more than 101 redundant parts to their unmanned company's processing infrastructure. 101 is plenty, maybe too many, for the average owner to keep track of how they are performing.  Adding more parts reduces the degree to which each part can be vetted and therefore reduces the system's reliability.   Total reliability is a combination of node redundancy and node reliability via reputation-based vetting.

In BitShares, absolute control is fully decentralized down to the votes of every single atomic BTS satoshi.  You can't get more decentralized than that.
 
EDIT:  I have updated the above post because I get smarter (and better looking) each day.  
My thanks to my fencing partners here who have helped me to hone my explanation with each post.

[Daedelus]

It is crazy to me how people focus on the block producer cap instead of the underlying consensus metric. If you remove the cap and let people produce proportionally to their approval, you would see a higher concentration of blocks produced by a smaller number of nodes.

Having 101 validators is FORCING the network to be MORE decentralized than the other consensus protocols, where 99% of blocks are produced by far fewer than 100 validators. As a stakeholder, your control over block producers is exactly proportional to your stake whereas in other systems it becomes unprofitable to validate unless you control huge amounts of the stake (or hash power in POW, or unique nodes on ripple UNL)

What is the goal of decentralization? It is to make it difficult for anyone but majority stake-vote to decide what the rules of ownership are. To protect the stakeholders from everyone else. The goal is not to make everyone feel like a part of the validation process and pay them for make-work.

You are talking about the situation as it is today. 101 nodes might be acceptable to some for the amount of users in the network today.


But what happens if node growth lags (way) behind network growth (which is likely do to the economic incentives of the node operators). Is 101 nodes for the economy of 3 billion people decentralised?

Rather than enforcing decentralisation, the fundamental concept of capping participation enforces centralisation over time as the network expands.

The ratio between node operator and user can only have an upward trajectory in a capped system;

1:10 > 1:50 > 1:200 > 1:1000 .....     > 1:1Million


This pattern is growing centralisation. A system destined for growing centralisation can't be called decentralised. Where is the logic?


You need to show why Bitshares won't follow this path.

[fluxer555]

A system destined for growing centralisation can't be called decentralised. Where is the logic?

Your statement is at odds with itself:

1. By describing "A system destined for growing centralisation" you acknowledge that there many shades/degrees of centralization/decentralization.
2. By stating "...can't be called decentralised", you force a description of said shades/degrees into a binary TRUE/FALSE state.

That being said, I think the point you bring up about increasing centralization with network growth is a valid concern, and is a useful contribution to continuously-improving BitShares hivethink.

[Daedelus]

There are shades, the main three distinctions are centralised > distributed > decentralised with shades in each.

I keep seeing arguments framing everything in terms of decentralisation. I read stan's 7 ideas as an example of it and it is absurd.  

Imagine starting with the colour red. Add a bit of yellow an it is a different colour. Keep adding yellow and there will be a point were you can't call it red any more and soon it will clearly be orange in 95% of people's eyes. Stan is starting with orange and trying to make us think of orange only in terms of red. It doesn't work.

We already have a word to describe this setup, it is distributed. Attempts are being made to twist this into 'sufficiently decentralised', attempting to pass a clear orange in 95% of people's eyes off as the reddest of reds (most decentralised).  


Decentralisation is a very special word is crytpo, it is a foundation concept and core principle so its definition shouldn't be diluted.

[toast]

Decentralization is about who is in control.

The original design was pure TAPOS with a *single* block signer! This would still be decentralized by our interpretation because SHAREHOLDERS decide who that block signer is and because the block signer does not have the ability to do any damage if clients refuse to roll back. The problem with this design is that it is a million times easier to coordinate the transition between block signers on-chain than off-chain, so we want enough block signers to stop them from coordinating to block all transactions and tada you've invented DPOS.

Who cares about the ratio of users to block producers? Block producers have no power and serve the BTS holders. Again, what is your goal?

Our goal is to protect BTS (and by extension, bitAsset) holders from non-BTS holders. That's it.

[toast]

You are talking about the situation as it is today. 101 nodes might be acceptable to some for the amount of users in the network today.

To the contrary, RIGHT NOW all other systems are more decentralized because they are small enough to afford more than 100 validators. DPOS will only become more decentralized AT SCALE, when you cannot afford that many validators unless you force them to be distinct. The argument is presented in depth in a few places:

http://bytemaster.bitshares.org/article/2015/01/07/The-Most-Decentralized-Proof-of-Stake-System/
http://wiki.bitshares.org/index.php/DPOS#Scalability

But as I said in my previous post, focusing on the ratio of block validators to users is a red herring. Focus on the ratio of stake in the network to influence over network behavior. Other POS systems force out the little guy.