10 ms·

How do layer 2s differ from execution sharding?

90 points by bpierre 2y ago

polyomino 2y ago
As far as I’m aware, there’s no way of validating something without observing it. If that’s true, then you will have to trade the ability to validate the entire chain in order to scale on chain, at which point, who cares what’s on chain vs off?
- yokem55 2y ago
  > As far as I’m aware, there’s no way of validating something without observing it. That's where zero knowledge proofs come in. If you know the root hash at state a, and the new root hash at state b is given along with a zk proof, that proof data can quickly prove that the state change from a to b was arrived at correctly. In this case, the benefit of doing this on ethereum (or other evm l1 blockchain) is that those proofs can be validated on chain.
- _reza 2y ago
  ZKP and FHE let you verify whatever you like. The former leaks data while the latter doesn't.
  - matthewdgreen 2y ago
    Can you explain what you mean?
jongjong 2y ago
This is a good article. With Optimism, Ethereum actually looks like a complete, usable solution that can scale. That said there is a lot of complexity around bridging which leads to high transaction fees, long waiting times (e.g. some projects have a 7+ days waiting time for bridging tokens back from L2 to Ethereum mainnet). Another downside to Ethereum is that it forces all this legacy on everyone... As many L2s and other ERC20 tokens will be abandoned in 20+ years, much of their data will still be stored on Ethereum mainnet blockchain. Smart contacts are more like accounts, they aren't like old transactions, you can't just prune them after some time without major ramifications. With old transactions in single-token blockchains, you can potentially prune old transactions while keeping account data and balances the same. The trust and centralization implications of doing this are minimal and constrained to that single blockchain only. With Ethereum, Ethereum community can't just decide to prune old unused smart contracts from the mainchain since it would affect the integrity of L2s and it's noth Ethereum's place to decide whether or not they can delete tokens that belong to other communities. Overall, Ethereum ecosystem has a legacy problem due to the fact that they are multi-token. L2s accumulate unnecessary legacy from Ethereum and also from other L2s. IMO, the benefits of decoupling tokens from Blockchain infrastructure aren't worth it as the value of a token is ultimately derived entirely from its infrastructure/integrations.
- b112 2y ago
  Compute power marches on, including bandwidth, i/o, and storage space/speed. If it increases faster than blockchain size, then we're OK.
  - ur-whale 2y ago
    > If it increases faster than blockchain size, then we're OK. Hope is not a strategy, and things should actually be the other way round: to avoid centralization, blockchain projects should be designed to scale at the same rate available compute power does externally: b/w, storage, etc.
    - tossandthrow 2y ago
      It absolutely is. Everything around you banks on an ever expanding economy and continuing inflation. Housing, pensions, etc. Saying that this is not a strategy is just plain wrong.
  - lttlrck 2y ago
    So long as we can pump oil out of the ground faster than we use, we're ok. A surprising number of things are "OK" if we ignore externalities.
    - b112 2y ago
      This is flawed reasoning. I mentioned nothing about oil, you are bizarrely trying to attribute a techinal response, to an anti-environment response. You fo realize ethereum isn't bitcoin, yes? Regardless, it sounds as if you're against more efficient, power saving computing? (Hands down, compute power is incredibly power efficient, compared to 20 years ago. This trend does, and will continue.) If you have issues with ethereum, which is one of the most friendly power consumption cryptos which exists, don't blame people citing techical facts.
  - jongjong 2y ago
    It doesn't solve the centralization issue though since it becomes increasingly difficult to run a full 'archive node' over time. I suspect it would take months to sync a full node from scratch today. That barrier is too high to be called decentralized. You end up having to restore from snapshots and that involves trust.
CynicusRex 2y ago
[flagged]
lawrenceyan 2y ago
I would like to point out that the alternative method of focusing on maximizing execution bandwidth and speed at the base layer is being done on Solana, with great success. Layer 2s if they’re needed are as easily added on top as it is for Ethereum. But Solana is so fast and cheap already right now that it hasn’t been necessary. I’m of the philosophy that since you only get to build one base layer, you should really spend the time to do it well and have it designed to be as efficient and performant as possible.
- Hendrikto 2y ago
  Solana is certainly fast and cheap, but to the detriment of reliability and security. Not sure that this is the correct tradeoff.
  - lawrenceyan 2y ago
    Ethereum and Solana are exactly the same in reliability and security. But Solana is orders of magnitude faster and cheaper at the same time. This wasn’t easily achieved by the way. It’s a heroic feat of engineering that shouldn’t be discounted.
    - curzondax 2y ago
      what are you smoking? Solana has had various outages in its lifetime and required to repeat transactions a lot. thats NOT the same experience as on Eth and far inferior in reliability.
      - davidmurdoch 2y ago
        They're just trying to "pump their bags", as they say.
    - arandomusername 2y ago
      By increasing the requirements to run a node by multiple factors, yeah. Solana recommends 10Gbit/s network speed for running a node. Ethereum nodes can easily run on almost any home connection.
jbglenn 2y ago
Kaspa solved the trilemma without sharding using DAGs. Nothing else can compete.
tansan 2y ago
What does this mean for L1? would it be less desirable to launch anything on L1 (ex, DeFi, NFTs, etc...)?
- arandomusername 2y ago
  L1 has higher liq and more high value users. L2 is mainly used for shitcoins & NFTs that will be worthless in couple of weeks/months
clpmsf 2y ago
From a high level engineering perspective (and not being an expert in blockchain systems), the amount of complexity in the Ethereum ecosystem seems... excessive? And it seems likely the amount of complexity will only continue to increase over time. We tend to prefer 'elegant solutions' and simple systems in engineering, and this seems quite the opposite. Is it truly necessary to solve the problems that it's trying to solve?
- castalian 2y ago
  Specific decisions or the need to preserve legacy may be questionable, but overall - yes, blockchains fundamentally don't scale well.
- whimsicalism 2y ago
  I'm not sure how much more 'actually complex' it is. I think we can often underestimate the complexity of incumbent systems because most of the complexity is not in the open. I'm sure that market makers and exchanges in traditional finance have a lot of complexity behind them.

90 points by bpierre 2y ago

As far as I’m aware, there’s no way of validating something without observing it. If that’s true, then you will have to trade the ability to validate the entire chain in order to scale on chain, at which point, who cares what’s on chain vs off?

yokem55 2y ago
> As far as I’m aware, there’s no way of validating something without observing it. That's where zero knowledge proofs come in. If you know the root hash at state a, and the new root hash at state b is given along with a zk proof, that proof data can quickly prove that the state change from a to b was arrived at correctly. In this case, the benefit of doing this on ethereum (or other evm l1 blockchain) is that those proofs can be validated on chain.
_reza 2y ago
ZKP and FHE let you verify whatever you like. The former leaks data while the latter doesn't.
- matthewdgreen 2y ago
  Can you explain what you mean?

jongjong 2y ago

This is a good article. With Optimism, Ethereum actually looks like a complete, usable solution that can scale. That said there is a lot of complexity around bridging which leads to high transaction fees, long waiting times (e.g. some projects have a 7+ days waiting time for bridging tokens back from L2 to Ethereum mainnet). Another downside to Ethereum is that it forces all this legacy on everyone... As many L2s and other ERC20 tokens will be abandoned in 20+ years, much of their data will still be stored on Ethereum mainnet blockchain. Smart contacts are more like accounts, they aren't like old transactions, you can't just prune them after some time without major ramifications. With old transactions in single-token blockchains, you can potentially prune old transactions while keeping account data and balances the same. The trust and centralization implications of doing this are minimal and constrained to that single blockchain only. With Ethereum, Ethereum community can't just decide to prune old unused smart contracts from the mainchain since it would affect the integrity of L2s and it's noth Ethereum's place to decide whether or not they can delete tokens that belong to other communities. Overall, Ethereum ecosystem has a legacy problem due to the fact that they are multi-token. L2s accumulate unnecessary legacy from Ethereum and also from other L2s. IMO, the benefits of decoupling tokens from Blockchain infrastructure aren't worth it as the value of a token is ultimately derived entirely from its infrastructure/integrations.

b112 2y ago
Compute power marches on, including bandwidth, i/o, and storage space/speed. If it increases faster than blockchain size, then we're OK.
- ur-whale 2y ago
  > If it increases faster than blockchain size, then we're OK. Hope is not a strategy, and things should actually be the other way round: to avoid centralization, blockchain projects should be designed to scale at the same rate available compute power does externally: b/w, storage, etc.
  - tossandthrow 2y ago
    It absolutely is. Everything around you banks on an ever expanding economy and continuing inflation. Housing, pensions, etc. Saying that this is not a strategy is just plain wrong.
- lttlrck 2y ago
  So long as we can pump oil out of the ground faster than we use, we're ok. A surprising number of things are "OK" if we ignore externalities.
  - b112 2y ago
    This is flawed reasoning. I mentioned nothing about oil, you are bizarrely trying to attribute a techinal response, to an anti-environment response. You fo realize ethereum isn't bitcoin, yes? Regardless, it sounds as if you're against more efficient, power saving computing? (Hands down, compute power is incredibly power efficient, compared to 20 years ago. This trend does, and will continue.) If you have issues with ethereum, which is one of the most friendly power consumption cryptos which exists, don't blame people citing techical facts.
- jongjong 2y ago
  It doesn't solve the centralization issue though since it becomes increasingly difficult to run a full 'archive node' over time. I suspect it would take months to sync a full node from scratch today. That barrier is too high to be called decentralized. You end up having to restore from snapshots and that involves trust.

CynicusRex 2y ago

[flagged]

lawrenceyan 2y ago

I would like to point out that the alternative method of focusing on maximizing execution bandwidth and speed at the base layer is being done on Solana, with great success. Layer 2s if they’re needed are as easily added on top as it is for Ethereum. But Solana is so fast and cheap already right now that it hasn’t been necessary. I’m of the philosophy that since you only get to build one base layer, you should really spend the time to do it well and have it designed to be as efficient and performant as possible.

Hendrikto 2y ago
Solana is certainly fast and cheap, but to the detriment of reliability and security. Not sure that this is the correct tradeoff.
- lawrenceyan 2y ago
  Ethereum and Solana are exactly the same in reliability and security. But Solana is orders of magnitude faster and cheaper at the same time. This wasn’t easily achieved by the way. It’s a heroic feat of engineering that shouldn’t be discounted.
  - curzondax 2y ago
    what are you smoking? Solana has had various outages in its lifetime and required to repeat transactions a lot. thats NOT the same experience as on Eth and far inferior in reliability.
    - davidmurdoch 2y ago
      They're just trying to "pump their bags", as they say.
  - arandomusername 2y ago
    By increasing the requirements to run a node by multiple factors, yeah. Solana recommends 10Gbit/s network speed for running a node. Ethereum nodes can easily run on almost any home connection.

jbglenn 2y ago

Kaspa solved the trilemma without sharding using DAGs. Nothing else can compete.

tansan 2y ago

What does this mean for L1? would it be less desirable to launch anything on L1 (ex, DeFi, NFTs, etc...)?

arandomusername 2y ago
L1 has higher liq and more high value users. L2 is mainly used for shitcoins & NFTs that will be worthless in couple of weeks/months

clpmsf 2y ago

From a high level engineering perspective (and not being an expert in blockchain systems), the amount of complexity in the Ethereum ecosystem seems... excessive? And it seems likely the amount of complexity will only continue to increase over time. We tend to prefer 'elegant solutions' and simple systems in engineering, and this seems quite the opposite. Is it truly necessary to solve the problems that it's trying to solve?

castalian 2y ago
Specific decisions or the need to preserve legacy may be questionable, but overall - yes, blockchains fundamentally don't scale well.
whimsicalism 2y ago
I'm not sure how much more 'actually complex' it is. I think we can often underestimate the complexity of incumbent systems because most of the complexity is not in the open. I'm sure that market makers and exchanges in traditional finance have a lot of complexity behind them.