Rollup & L2 Classification on L2Fees.info

Background

L2Fees.info aims to display an apples-to-apples comparison of the fees for users to take common actions on Ethereum scaling solutions. While fee comparisons across layer-1 blockchains must account for the security and decentralization tradeoffs made by various blockchains, L2Fees only compares chains that inherit Ethereum’s security.

As well-respected leaders in the layer-2 space, we’ve delegated the curation and classification of projects listed on L2Fees to L2Beat.com.

Recently, the Metis project adjusted their data-availabilty system, leading to community discussion around the security assumptions of the chain. L2Beat.com reclassified the project as an Optimistic Chain, and community members began pushing for the project to be reclassified on L2Fees as well:

The site was updated to move Metis to a newly created “All L2s” tab, however the Metis team has pushed back against this change.

Goal

In light of this discussion, I’d like to use this opportunity to engage with the community, and establish the following properties in relation to L2Fees.info:

  1. What qualifications are required for a project to be listed in the CryptoStats l2-fees collection and on L2Fees.info
  2. How should L2Fees.info present data that is simple, accurate and informative

Current status

I’d like to share my current take on these questions, and invite the community to respond to them:

The l2-fees collection collection includes protocols that fit L2Beat.com’s definition of “a chain that fully or partially derives its security from L1 Ethereum so that users do not have to rely on the honesty of L2 validators for the security of their funds”.

Given that data availability is currently the most expensive resource for rollups, many projects have moved their data off of Ethereum, such as Metis (classified by L2Beat as an “optimistic chain”), ImmutableX, DeverseFi and Sorare (all classified as Validiums).

Given that this is a significant change in the security assumptions of an L2, L2Fees should clearly differentiate these projects from full L2s. This is currently accomplished by having two tabs: “Rollups” and “All L2s”.


I look forward to seeing the community’s thoughts on this topic!

3 Likes

My 2 cents that hiding Metis in the separate tab that is not even obvious for the end user is not the solution for the “security concerns” or for “fees too low to be secure” reasons. Lets resolve the immediate problem with this hiding, un-hide Metis and then get into long discussion about technology. I am here to answer the questions and also invited our tech experts to chime in. Metis is L2, rolling up transactions, saving them on L1 , but in a SMART way - loading on demand. Metis is very first in the industry that came up with this tech, Arbitrum will be next to do the same - we discussed that with their team already. We also are in talks with Ethereum Foundation on the classification of this brand new approach.
No one in the industry has done it, this is completely new, this is pure innovation.
Lets not throw out the baby of innovation with the water of the old thinking.

1 Like

Creating apples-to-apples comparison is very hard as the security assumptions of all Layer2s listed on l2beat.com are slightly different (hence the detailed risk view trying to describe these nuances). That’s not to mention different scaling solutions that do not derive their security from mainnet Ethereum at all.

Rollups - ie systems that use Ethereum for both data availability AND to check validity of the state (either through validity proofs or fault proofs) are considered to be systems that derive their security from Ethereum. Ignoring admin upgrade keys they can be considered as secure as Ethereum if properly implemented.

If a system does not put data on L1 Ethreum it cannot be considered a Rollup - it does not matter if data is put into the external system (as with Validiums) or is assumed to be available “on-demand” as in the case of Metis. Clearly there are different security assumptions.

What are these security assumptions ? With Validiums as implemented today by Starkware it is simple - at least one member of DAC must be honest. With Metis it is more nuanced - users must trust the cryptoeconomic setup of the whole system that will incentivise honest Validators to force Sequencer to release the data during potential data witholding attack. The discussion of whether this is secure mechanism is complex and highly dependend on the details of the current implementation and actual parameters such as bond size, etc… Worst case scenario with Metis today is that the security falls back on the honesty of Validators to remove dishonest Sequencer, i.e. the typical security assumption of the sidechain.

How the l2fees.info should present the information is a very debatable topic of course but in my view users should be aware that main cost factor for rollups (a security tax) is a calldata cost and having Rollups and Metis on the same list should be done so it is clear which systems are Rollups (i.e. they derive their security from Ethereum) and which have additional security assumptions.

Side note - while Arbitrum indeed plans to implement “AnyTrust” chains that will not put data on-chain, however their additional trust assumption is that of a DAC, similar to Validium. That is a very different security model than Metis.

1 Like

I don’t think they are hiding Metis in a separate tab. Metis is classified different now. Metis is storing their fraud-proof data on a decentralized blockchain called “Memo labs”. This will lead to cheaper transactions. But if the sequencer needs to check / or call for the fraud proof, while the “third party blockchain” Memo labs is under a 51% attack or down, how will metis know that the data they are calling is still valid? If this stored data gets compromised, it’s impossible to verify the blockchain and there is no recovery? I think this is why the Metis is not a rollup anymore and this is why you should disclaim that to your users. The trust assumption is not Ethereum anymore. It’s a huge trade-off and not worth it, with upcoming EIP improvements.

On a side node, you are the CEO of Metis. My 2 cents is that you care more about marketing cheaper transactions than you care about the safety of user funds. At least you can let the user choose for themself if they want Ethereum as a security model or Memo Labs…

2 Likes

You are missing the point of the argument.

  1. Metis is different from legacy Rollup, but it still relies on Ethereum to do the finality and fraud proof. We are still paying the L1 security fees. Metis is a Layer2. Find more details here https://metisdao.medium.com/the-tech-journey-lower-gas-costs-storage-layer-on-metis-867ddcf6d381
  2. This website is called l2fees.info, so it is l2fees, not just rollup fees, right? What we are suggesting is that the website should list all L2s as the default page, instead of just rollups, then you can have more categories to list different L2 scaling solutions.
  3. Assuming Metis is not caring about the safety of the network is baseless. Meanwhile, attacking the CEO of Metis is not a gentleman should do and is violating the guidelines of this forum!
2 Likes

Hey David! I agree with your approach of including only chains that fully or partially derive their security from L1 Ethereum.

Let me explain how Metis Andromeda derives its security from L1 Ethereum:

  1. Metis L2 collects/rolls up a batch of txs
  2. Tx data is posted to Memolabs with its Merkle Tree Roots, and the Tx Batch MTR & State Roots are posted on L1 (Transaction commit submitted to L1)
  3. Whenever needed (fraud proof posted / data needing to be verified on L1): the rest of the data gets brought back on-chain via the MTR (used as an identifier tag) and everything is resolved in Layer 1. (L1 validators can ask for a resubmission of the data to L1 in order to make sure it matches the originally submitted commit)

If the request is not honored, L2 sequencer gets slashed.

“What about the fisherman dilemma tho?”
Metis avoids the fisherman’s dilemma by making the sequener role not permanent. In other words, the cost of kicking out a bad sequencer is finite

People often get confused thinking Metis has its own consensus mechanism when hearing about the peer network, but this assumption is very wrong. Metis’s peer network is there to catch production issues sooner, but that is nothing more than an extra security layer: it doesn’t replace Ethereum’s consensus mechanism, it adds to it:slight_smile:

Responding to point #2: Nothing wrong with the new “Rollup” and “All L2s” tabs recently created. The issue lies in the fact that instead of having “All L2s” as the landing page for L2fees, the organization chose the “Rollup” tab as the landing page, which looks like a subtle attempt to hide Metis/make it look like its not an L2, while showcasing the VC-backed ones.

If someone enters L2fees and see a bunch of L2s, but not Metis: they’ll think Metis is not an L2.

It’s simple.

It’s fixable.

Should be fixed:)

2 Likes

I don’t have much to add to the discussion above. I just want to voice my opinion on how this could be resolved. The cryptofees.info website should by default show all L2s, but there should be a visible tag to indicate what category of L2 it falls into. I believe that as more and more solutions get added to the dashboard like validiums, volitions, Arbitrum AnyTrust, optimistic chains, etc, we should have a simple legend that shows which color or symbol is which and add that next to each item on the list to help categorize them in the All L2s view page. Other than that I agree with the opinions voiced above: 1. Metis should not be in the rollup section and rightfully so, 2. It should not be hidden from the default view which should be all L2s. 3. The All L2s view should have the classifier legend I mention that indicates security assumptions.

1 Like

100% agree with your first 2 points. Although, it’s important to remark that Metis Smart L2 does rely on Ethereum L1’s underlying security. Correct me if I’m wrong, but Validiums don’t.

I believe Validiums and volitions can be added as well. Nevertheless, even if “L2 solutions” not relying on L1’s security are not added to L2fees, Metis should still be added.

Just want to emphasize that Metis shouldn’t be compared to Validium, as Metis does support general computations and does rely on Ethereum L1’s underlying security (Key feature for a protocol to be considered an L2).

2 Likes

Validiums do not inherit L1 security, correct, they have off-chain data availability. So if I understand correctly, Metis inherits Ethereum security if and only if, the prover on L1 requests the state, the sequencer needs to make that data available otherwise it gets slashed. Henceforth the ‘optimistic L2 chain’ terminology. It inherits security optimistically (when challenged)? I thought that to inherit security all calldata and state need to be posted to L1, I still haven’t fully understood how the current mechanism makes Metis inherit security. On another note; once blob storage is available post EIP-4844, are you going to start posting calldata to blob storage and become a rollup? Or is the MEMO storage architecture decision a more permanent one?

Exactly. Metis inherits security from Ethereum L1 by storing the MTR and the State Roots on-chain. If any tx was to be altered, the MTR stored on chain would not match the MTR of the transaction data stored on Memolabs. Because of this, there’s ultimately no way no way for the “fradulent data” to go past Ethereum’s L1 security layer. Fradulent data would have to get past Metis peer network first, but if achieved: Ethereum’s L1 security is there to save the day.

On your second point, yes. We’re going back to pure Optimistic Rollups post sharding:). We’re going to be able to process 100,000 TPS, while leaving behind the debate about if Metis does or doesn’t inherit L1’s underlying security. Very exciting things are coming.

2 Likes

I don’t have much to add - only that the argument seems to hinge on how much (and in what ways) a protocol must rely on the underlying L1 security to be considered an L2.

It’s not arguable that Metis relies on the L1 security. It’s also not arguable Metis does not have it’s own consensus mechanism.

These 2 facts alone make me think it deserves a spot that is just as visible/prominent as other L2 solutions. If the site wants to sub-divide L2s into unique categories - this seems appropriate.

However - unless the site has a formalized definition around the type of security requirements between the L1 and L2 - it doesn’t seem prudent to kick protocols to less visited areas of the site. Perhaps the community should be asked to create a more formal/rigorous definition before they are asked to see if certain protocols meet a sorta non-sufficient standard.

Anyway - I appreciate the thoughtful conversation here. I also appreciate not being afraid to think smartly about solving the Blockchain trilema. Excited to see where the Metis “Smart L2” thinking lands within this community.

2 Likes

while the “third party blockchain” Memo labs is under a 51% attack or down, how will metis know that the data they are calling is still valid?

The Sequencer is obligated to provide the data on-chain upon request in that situation. Memolabs is used so that Validators don’t have to request data directly from the Sequencer every time. If Memolabs is down/attacked, the network is still secured via data requests.

1 Like

I value security over anything. Nothing matters more to me. If you met or know me - you know that. I went through this design back and forth and have a strong confidence that L2 model that Metis adopted is as secure as a traditional Optimistic Rollup. We are working on the additional documentation and articles to explain all the details and this forum helped us with what exact areas we need to document more.

The concern that Metis was singled out and was hidden from the L2FEES landing page is STILL NOT ADDRESSED, so please address and fix that part!

The Sequencer is obligated to provide the data on chain upon request in that situation. Memolabs is only a decentralized storage so that Validators do not have to request data every time. If it is down, the network is still secured via data requests.

The mechanism to keep the sequencer in check is also mentioned by John Adler in the “On Data Availability Challenges” section. Metis resolves some of the main criticisms in the “A game of cat and mouse” section.

Agree with Dcbuilder’s recommendations.

Since the L2fees website is dedicated to L2s, the landing page should reflect that and not land on Rollups views. Descriptors can be added to further categorize L2s.

We all came here to have a productive discussion, and we are looking forward to hearing l2fees’ team feedback.

Thanks to everyone for their detailed and thought-out responses. Sorry it’s taken me a bit to respond, but I didn’t want to rush a response, I wanted to give a full read to every thing written here.

I certainly understand Metis’s system much more than previously. However, I’m still no expert on L2s, and I feel that, until L2Fees can have a well-defined criteria for listing on the site, it’s best for us to delegate that responsibility to our friends at L2Beat.

Furthermore, I think @dcbuilder’s suggestion is probably the most reasonable approach for how to present data on the site: show all protocols that are classified as L2s by L2Beat, but provide a visual indication for which ones are “rollups” and which are “other L2s”. This will probably come in the form of tooltips (similar to the existing tooltips for “throttled” chains), as well as some color differences.

I’ll talk to some of the designers in our community to get their thoughts on how to best represent it, but feel free to leave any thoughts you guys may have on how this would be best represented.