What Are Ethereum Layer 3’s?

Ethereum Layer 2’s are gaining traction: they save users transaction fees and help create space on the Layer 1, the Ethereum blockchain. A third layer would partly be about scaling: allowing Ethereum to become that world computer, capable of transacting limitless amounts of transactions. But a third layer would also be about customized functionality. Some applications need specific tailoring that may better be served by a separate layer. And lastly, layer 3 solutions will allow small ecosystems within Layer 2’s to grow without paying a ton of gas fees.

Layer 3 development is still in an early stage, and highly conceptual!

Breaking Down the Stack: Layer 1, 2 & 3

There’s a lot of different ways in which the terms Layer 1, 2 and 3 are used. Let’s outline the stack and clarify how we DON’T use the term Layer 3 in this article.

Layer 1’s and 2’s

People agree more on what Layer 1’s and 2’s are than on what Layer 3 is. Layer 1 is the base layer, the underlying blockchain. On top of it, on – you guessed it – Layer 2, rests a separate blockchain. For Ethereum, there are several options. Ethereum Layer 2’s aim to increase the main chain’s transaction throughput by offloading a chunk of the transactions to Layer 2. Then, they send a cryptographic proof of the validity of the batch of transactions to Layer 1.

Why Layer 2’s? The increased cost of transactions on Ethereum Layer 1 is a problem, especially in bull markets 🙂 In those periods, the Ethereum Layer 1 throughput of about 15 transactions per second creates a bidding war for transactions to go through. In November 2021, the peak of the bull market, users paid around 50 dollars in gas fees for a token swap.

This is driving people to Layer 2’s such as Polygon. Ethereum stays the bedrock: it becomes a settlement layer for L2. It’s very possible that soon, end-users will be mostly active on the cheap L2’s. There’s more elbow room there, and life is cheaper.

So What is Layer 3? 

At the point where we reach Layer 3, some confusion about the meaning of the term can creep in.

• Layer 3 is sometimes used as a term for the application layer that uses Layer 2’s. To take an example from Bitcoin’s ecosystem: apps built on top of Lightning that allow you to authenticate yourself with a Lightning wallet. For Ethereum, apps like Uniswap are sometimes called Layer 3’s. But Layer 3 is sometimes also used to describe custodial services like centralized exchanges, such as Binance.
• Another way that people view Layer 3 is as a layer that allows for interoperability between blockchains, connecting the web 3 ecosystem.             

What do we mean here by Ethereum Layer 3?

So how do we use Layer 3 in this article, where we talk about Ethereum? In the case of Ethereum, Layer 3 is connected – through Layer 2 – to the base chain. It’s all software talking to each other, in a hierarchy of layers. Layer 3 talks to Layer 2 in the same way that Layer 2 talks to Layer 1. So these layers are more intimately functionally connected than when someone would say ‘payment provider Cash App is a Layer 3 solution for Bitcoin’.

The Limits of Data Compression

Some Ethereum Layer 2’s use so-called roll-ups, which is a cute way of saying data compression. To give an example from the real world: writing. You can reduce the number of space a written sentence takes up by leaving out the vowels. 

Fr xmpl, ths sntnce s stll (srt f) rdble.

In the above example, leaving out the vowels is the data compression: the roll-up. By the way, roll-ups on Ethereum can achieve an impressive 8x data reduction size. Impressive at least compared to the above sentence, which uses 41 characters compared to the 55 it would take for the normal spelling: only roughly 1.25x data compression.

But, as Ethereum founder Vitalik Buterin pointed out: ‘you can’t just keep scaling by stacking rollups on top of rollups’. Our above example of the sentence without vowels illustrates this: we can’t do the trick of removing vowels a second time. Data can be compressed once, but not again and again.

This is all to say that a Layer 3 isn’t necessarily about putting the same data rollup architecture on top of other roll-ups. Still, it can work for computation scalability, as we will see.

Hyper Scaling on Layer 3 Protocols

One idea for a Layer 3 would be to just stack a similar scaling solution on Layer 2. As discussed above in the discussion of roll-ups, this won’t always work. At least not for data compression.

But for computation, it might work. Starkware, also mentioned in Vitalik’s article about Layer 3 scaling, is one of the companies working on this.

Starkware is a company working on the bleeding edge of zero-knowledge technology. They have done promising experiments with recursive zero-knowledge proofs. Those would be zero-knowlegde proofs of zero-knowlegde proofs of… z.. you get it. 

Unlike the above example of data roll-ups, these can be scaled. For example, you can take 10 zero-knowledge proofs — each representing a thousand transactions — and generate an entirely new zero-knowledge proof verifying that those other 10 proofs are correct. 

To clarify, the following example. Suppose I want to prove to you that I spent 7 successive days in a certain location. On the first day, I take a selfie with the newspaper of that day. On day two, I make a selfie of myself holding both that day’s newspaper AND the selfie from day one… etcetera. 

Recursive proof in one photo: Source: Coingeek

This form of extreme scaling is achieved by the recursive pattern through the aggregation of proofs from L3, to be verified by a single proof on L2. StarkWare has already rolled up as many as 600,000 NFT mints into a single transaction on Layer 2 ImmutableX.

Besides scaling, layer 3 could have different purposes from layer 2. Vitalik lists as examples customized applications that don’t want to use the Ethereum Virtual Machine (EVM) to do computation. Another purpose of layer 3 would be supporting specific functionality, for example privacy tools.

Why a Layer 3?

But why even a third layer for these use cases? In other words, why anchor these new data processing tools to layer 2 instead of directly to layer 1? (see below schedule for these two alternative approaches)

Image source: Vitalik Buterin: What kind of Layer 3s make sense?One possible argument for the three-layer model is, according to Buterin, that it solves the tradeoff between confirmation time and cost for roll-ups and zk-proofs with relatively low volumes. A roll-up pays a high, fixed gas price for posting a batch on Layer 1. This is only efficient if this cost is shared by a whole lot of transactions in this roll-up. And in a future world with application specific environments that are rather niche or early in their adoption, the number of transactions could be too low to be profitable for Layer 2 rollups who have to use the expensive Layer 1. Why? Because it would become either too costly or the confirmation times would get too long. The interaction between this application specific environment on Layer 3 and Layer 2 wouldn’t have this problem: it’s not costly. Hence the benefit of an extra layer.