Scalability Tradeoffs: Why "The Ethereum Killer" Hasn't Arrived Yet


3 年以前

I've seen a lot of crypto-enthusiasts on Reddit and Telegram making comments like:
"Bitcoin is slow. It is expensive. There are many new coins, modern ones that are much better. They are fast and inexpensive."
Or the very popular CryptoKitties argument:
"Ethereum couldn't even handle CryptoKitties, how do you expect it to be Web3.0?"
Or about how Blockhain X is here to turn the tables:
"<insert coin ticker> is king, it can handle 60,000 transactions per second, has no fees and it can do smart contracts"
The popular opinion is that the current leaders by market cap are not good enough, and that new projects are offering better features or alternative architectures (Tangle, Hashgraph) that are going to define a new standard and bring the capabilities of blockchains to new levels.

While I do not dismiss the possibility of Bitcoin being dethroned in the upcoming years, or that the top-5 might change radically in the future, I believe that we need to be skeptical when a project advertises itself as a do-it-all solution, and rigorously investigate it before jumping to conclusions.

There is no silver-bullet that will solve all problems.

"Touka Koukan" (等価交換) is a Japanese phrase which roughly translates to "equivalent exchange". Nothing comes for free. There will always be trade offs.

Below is the Scalability Trilemma as described by Vitalik Buterin:


Α blockchain that claims to have solved the trilemma has either bent the laws of physics (highly unlikely), or it has discovered a breakthrough method that solves the major blockchain scalability problems that have stumped top mathematicians and computer scientists for the past decade.

While this is not impossible, a more likely explanation is that the blockchain has sacrificed either decentralization, security, or both.

What characterizes a blockchain or a cryptocurrency?

What characterizes a blockchain or a cryptocurrency?

In my debut article, A rant about Blockchains I provide the following definition of a blockchain:
A blockchain is a database that can be shared between a group of non-trusting individuals, without needing a central party to maintain the state of the database.
And cryptocurrency from Google dictionary:
A digital currency in which encryption techniques are used to regulate the generation of units of currency and verify the transfer of funds, operating independently of a central bank.
Note that both definitions (blockchain and cryptocurrency) emphasize the need to operate independently of a central party.

The case against (semi-)centralized coins


XRP claims to be "the fastest and most scalable digital asset, enabling real-time global payments anywhere in the world." They boast that "XRP consistently handles 1,500 transactions per second, 24x7, and can scale to handle the same throughput as Visa."

However, let's compare XRP's nodes number to Ethereum and Bitcoin:

425 Ripple Nodes around the globe |

11,690 Bitcoin Nodes |

32,383 Ethereum Nodes

In addition, Ripple can freeze its users' funds at any given time[1][2].

It may be fast, but Ripple is as centralized as it gets.

Masternode Coins

A Masternode is a full cryptocurrency node which can perform extra functions in its blockchain network such as:
  1. Participate in governance and voting
  2. Instant transactions
  3. Private transactions.
It is used in cryptocurrencies like DASH and PIVX.

In order to be able to run a masternode, you need to put up some amount of cryptocurrency as collateral — like stake in Proof of Stake — to disincentivize malicious behavior. In return, you get compensated with a cryptocurrency payout at regular time intervals.

The aforementioned features are great, however:
  1. A DASH masternode costs 1,000 DASH ($1,000,000)
  2. A PIVX masternode costs 10,000 PIVX (~$110,000)
The high financial barrier of entry translates to a certain degree of centralization which as discussed before, breaks the requirement of decentralization.
In order to maintain decentralization there needs to be a low barrier of entry for individuals to join and contribute to the network — This directly relates to the Bitcoin Civil war for the block size increase.

Permissioned or Private Blockchains

I will skip over this category completely since by definition this is a centralized setup. A highly recommended comparison on Public vs Private blockchains can be found here.
I also want to emphasize that a blockchain claiming to achieve a high number of transactions in a testnet does not mean that performance will translate if/when it is deployed as a public blockchain.

Zero / Ultra Low transaction fees

What if a coin claims to have zero transaction fees?

In order to keep the network secure, there is a need for verifying entities who create new blocks and maintain the network.
Why would these entities support the network and not try to attack it?
There needs to be an incentivizing method for them to do this task honestly without trying to game the network. That incentivizing method is transaction fees and block rewards. Since block rewards are slowly diminishing (e.g. block reward halving in Bitcoin) miner incentives are directly linked to transaction fees.
It is impossible to have a truly decentralized network with 0 fees as there would be no incentive to maintain it.

What about coins that have ultra low transaction fees?

I will rebut this argument with an example from Ethereum:

An Ether transfer costs 21,000 gas. Let's say you want your transaction to be confirmed fast, so after consulting you set the gas price at 50Gwei/gas, resulting in total transaction fees of 0.00105 Ether.

If you made the transaction in Jan 13, 2018 (Ξ1=$1,400) it would cost $1.47.

If you made the transaction in Jan 15, 2017 (Ξ1=$10) it would cost $0.01.

Transaction fees (given that the calculation method stays the same) are constant when denominated in the native currency (Ξ1=Ξ1 always). A multifold increase in a coin's price, implies a multifold increase in transaction fees when denominating in fiat currency.

When arguing that Coin X has $Y transaction fees when evaluated at $Z per coin, think about what Y will be if Z becomes 100*Z.

Corollary: Most coins that claim to be cheaper than Bitcoin or Ethereum can only do so because they have very low transaction volume.
The increase of a coin's price and transaction volume are positively correlated with an increase in that coin's transaction fees.

Consensus Algorithms

Both Bitcoin and Ethereum use Proof of Work, the most widely-used consensus algorithm since the inception of cryptocurrencies. If you are not familiar with how Proof of Work and consensus algorithms I refer you to my previous articles [1][2]

There have been many attempts at alternative Proofs of {Something} in order to tackle the deficiencies of PoW, most notably with Proof of Stake.

Taking Bitshares or Steem as an example, they both use Delegated Proof of Stake (DPoS) as their consensus algorithm (not to be confused with Cardano's DPoS algorithm).

Think of DPoS as representative democracy. Instead of voting directly on an issue (on a block), you delegate your voting power to trusted "delegates" who will use that power to take a decision (mine a block) for the masses. This in turn allows less overhead and enables much higher transaction throughput. However, it is less decentralized because there are only a handful of delegates who hold the majority of the decision-making power, and power is corruptible.

There are valid use cases for DPoS: Steemit, for example, is a successful platform with performance comparable to traditional web servers, despite running on its own blockchain. Most of its value transfers are micro-transactions, so sacrificing decentralization and security is a fair tradeoff to make in order to achieve such high levels of scalability and performance.

However, for a cryptocurrency storing large amounts of value, or a platform that handles smart contracts that must be trusted by third parties, it is far more secure to use a fully decentralized PoW chain.

At Loom Network, we think blockchains like this are best implemented as application-specific sidechains, where the sidechain security is maintained by the main chain.

What about Tangle or Hashgraph?


IOTA uses a totally different system from a blockchain, based on a Directed Acyclic Graph (DAG), called the Tangle.

Essentially the Tangle's transactions have no timestamps (which means you cannot do complex smart contracts on them which rely on X happening before Y) and they tackle the double spend problem by referencing two previously confirmed transactions. pg.10

This looked particularly promising until MIT Researchers found a critical vulnerability in the curl hash function that IOTA created. Additionally, the IOTA foundation (like Ripple), has also frozen user funds in the past.

Also since this is a fee-less system, as discussed before, there is no incentive to run a full node.


Hashgraph is also a relatively new concept which is described in depth here:

Demystifying Hashgraph: Benefits and Challenges

However, as said by the Hashgraph team in their Telegram channel, it is currently deployed only in a permissioned network. This again breaks the requirement for decentralization since it's private; however, it might have a valid use case in corporate environments. It is probably too early to know if it will find its way to public networks.


My previous arguments are aimed towards the politics and the proper design of community-driven blockchains.

I do not believe that XRP, Masternode coins and private blockchains are cursed and should be avoided. Each has its use case.

However, they are inherently more centralized and should not be mistaken for a truly decentralized, impossible-to-censor and unstoppable cryptocurrency.

Regarding new experimental algorithms and network topologies, I really hope they are able to sustain the scalability they claim they can achieve, without ultimately sacrificing decentralization or security.

It's too early to get overly excited about these platforms (and declare them the "Ethereum killer") until they have proven themselves at scale.

LOOM Dapps

More articles

1. ERC721x — A Smarter Token for the Future of Crypto Collectibles
2. Universal Transaction Signing: Seamless Layer 2 Dapp Scaling for Ethereum 🔐
3. Your CryptoKitty Isn't Forever — Why DApps Aren't as Decentralized as You Think