Ethereum Layer 2 Scaling Solutions: Rollups, Sidechains, Validiums

When bridging funds from L1 to L2, funds are sent to a bridge contract on Ethereum and locked there. The L2 network then looks at the contract state and mints equivalent funds on the L2. Withdrawal times vary: Optimistic rollups like Optimism have a 7-day withdrawal period, while zk rollups can be faster (hours to minutes). Side chains like Polygon POS have different security guarantees - if the side chain goes offline, funds may be stuck. The bridge process is decentralized and trustless, using call data to store L2 proofs on L1.

Wrapped assets in cross-chain bridges create significant security risks because users no longer hold the actual underlying assets. When bridging assets between chains, the destination contract controls the assets, meaning users carry risk perpetually—not just during the bridging process but forever as long as they hold the asset. This risk can materialize through various scenarios: bugs in the destination contract, malicious upgrades, or compromised messaging protocols. Since users cannot directly interact with their assets, they are vulnerable to any issues in the supply chain. This fundamental shift in the user's relationship with their assets represents a critical security concern in the interoperability space.

Bridge security is a critical priority in cross-chain infrastructure. Native bridges that use single-sided pools of the same asset on different chains offer lower security risks compared to traditional minting and wrapping bridges. These liquidity layers also enable faster transfers and cheaper costs, representing a fundamental trade-off in the bridging space that will evolve as the market matures.

Cross-chain bridges that transfer assets between blockchains (like wrapped BTC on Ethereum) introduce security risks that must be carefully evaluated. Unlike native assets on their home chains, bridged assets rely on the security of the bridge protocol itself. If a bridge is compromised or forks unexpectedly, the bridged assets may be lost or duplicated. This risk is particularly relevant for assets like stBTC, which represents a bridge-based version of Bitcoin rather than native BTC, requiring careful consideration of the underlying security architecture.

Bridging to a layer 2 has different risks than bridging to an alternative layer 1 (like Solana or Avalanche). Layer 2 bridges are trustless and immutable, working regardless of security dependencies on the chain itself. Alternative layer 1 bridges can be destroyed if security assumptions on the other chain fail. Layer 2 bridges are always a bridge of last resort—if the bridge goes down, the whole network goes down, making them less risky.

EIP-4844, known as Proto-Danksharding, is the most significant change in the Cancun-Deneb upgrade. This execution layer improvement is designed to dramatically reduce transaction costs for Layer 2 roll-up solutions like Optimism and Arbitrum. By implementing this change, Layer 2 networks will become significantly cheaper to use than they are currently. This upgrade is part of a broader ecosystem-wide effort to lower transaction fees across the Ethereum network, making decentralized applications more accessible and cost-effective for users.

EIP-4844 is an Ethereum upgrade scheduled for mid-March 2024 that introduces 'blobs' (temporary data containers) to reduce Layer 2 transaction costs by separating data publication from permanent storage, making Ethereum more competitive against alternative data availability layers like Celestia and EigenDA while enabling Layer 2 rollups to publish data more efficiently without permanently burdening the main network.

EIP-4844 is a major network upgrade for Ethereum that introduced 'blobs' - a special footprint on the main Ethereum chain where layer 2 transactions have a dedicated space. This upgrade allows more data to be stored in that space, effectively reducing transaction costs for layer 2 networks like Base. The video notes that transaction fees on Base have dropped to one cent or sub-cent levels, making it highly competitive with networks like Solana.

Proto-Danksharding (EIP-4844) is an Ethereum upgrade that dramatically reduces Layer 2 transaction costs by 50-100x, making decentralized finance more accessible while maintaining the security benefits of roll-ups through shared security with Ethereum and proof posting to Layer 1.

Proto-dank sharding is being proposed as a key first step toward full dank sharding implementation. It comes with Ethereum Improvement Proposal 4844, authored by Ethereum researcher Dank Rad Feist and co-founder Vitalik Buterin. The proposal involves adding a new transaction type called blob-carrying transactions to Ethereum. Currently, Ethereum blocks can carry 50-100 kilobytes of data, but under proto-dank sharding, they'll be able to carry closer to one megabyte—a 10 times increase in data availability. While this is less than the 16 megabytes projected under full dank sharding, it should significantly decrease the cost of using roll-ups and pave the way for cheaper, quicker layer 2 networks.

App chains represent a design philosophy where different applications use customized blockchain architectures suited to their specific requirements. The threshold for on-chain business remains expensive, creating a chicken-and-egg problem where high costs deter adoption while low adoption maintains high costs. L3 solutions aim to reduce data availability redundancy requirements, potentially bringing costs down to levels suitable for applications like social media and video streaming. Success requires timing the market when data availability costs remain sufficiently high to justify experimentation with alternative architectures.

Layer 2 solutions and App chains represent different approaches to blockchain scalability with distinct trade-offs. Layer 2 solutions inherit security from their underlying Layer 1 but are constrained by the Layer 1's architecture and cannot customize consensus mechanisms. App chains can customize the entire stack but must sacrifice some decentralization and security guarantees. Interoperability between chains varies: Layer 2 solutions like Optimism and Arbitrum cannot directly communicate at 100% trust level because they rely on the same underlying Layer 1. Cosmos chains can communicate directly through IBC because they share the same consensus mechanism. The choice between them depends on security requirements, customization needs, and target use case.

Appchains are specialized blockchains designed for specific use cases, representing a growing trend as applications migrate to this solution. Major projects like Uniswap and MakerDAO are developing their own appchains (Unichain and Layer 1 respectively) to consolidate their ecosystems. Cosmos implements a hub-and-spoke architecture with the Cosmos Hub connecting specialized blockchains. However, early adopters like DYDX faced challenges when migrating to Cosmos, as the ecosystem was losing momentum and the ATOM token did not capture value from applications. The key trade-off involves creating dedicated Layer 1s for complete control versus Layer 2 solutions that inherit underlying security. Appchains enable specialized risk management for markets like RWA, where traditional liquidity pools may not serve institutional needs adequately.

Appchains are specialized blockchains designed for specific applications that derive their security properties from underlying systems like Ethereum, while Layer 3 solutions focus on scaling and composability. The key distinction lies in security inheritance: appchains inherit security from their base layer, whereas Layer 3 solutions may offer more customization but require different trade-offs in terms of network quality units and centralization concerns. Adoption barriers include user experience complexity, regulatory challenges, and the need for abstraction layers to unify multi-chain interactions. ZK technology is predicted to dominate Layer 3 infrastructure, while appchains will continue to serve niche applications requiring specialized primitives like gaming systems.

This segment covers the emerging paradigm of Layer 3 architectures in blockchain development and the technical innovations enabling them. The speaker explains that Layer 3s represent a convergence of Layer 1 security and Layer 2 scalability, enabling appchains with custom consensus mechanisms and gas tokens. He argues that Layer 3s offer the 'best of both worlds' by allowing developers to customize protocols for specific use cases while inheriting security from Layer 1. The segment provides a deep dive into Cairo, a new programming language developed by Starkware for Starknet, and Kakarot, a Layer 3 project that implements an EVM-compatible virtual machine using Cairo. The speaker discusses how Layer 3s can offer custom consensus mechanisms like HotStuff from Aptos/Sui while maintaining Starknet compatibility. He mentions projects like Madara (using Substrate for Starknet sequencing) as examples of this emerging paradigm. The segment also covers the trade-offs between security and customization in Layer 3 architectures, noting that for applications like gaming or social media, the security requirements may be lower than for financial applications, making Layer 3s appropriate.

A fundamental challenge in L2 design is the centralization of sequencers, which contradicts Ethereum's decentralization ethos. Sequencers provide centralized performance benefits—faster confirmation, higher throughput, better UX—but introduce potential censorship risks. The escape hatch mechanism addresses this by ensuring L2 state is maintained on Ethereum, allowing users to always access their assets directly even if sequencers misbehave. This forces sequencers to behave properly because users can bypass them. The market appears to accept this trade-off, as evidenced by Base's rapid growth despite centralized sequencers. The escape hatch represents a fundamental safety mechanism that balances performance with decentralization.

ZK Sync currently has a centralized sequencer, which is a common practice in early-stage roll-up development for efficiency and speed. However, the system includes escape hatches that allow users to enforce any transaction and withdraw assets to Ethereum if they don't trust the network anymore. This is achieved by proving that any transaction call was made from Ethereum mainnet through a priority queue, and proving the outcome of that call. This design ensures that even if the sequencer becomes malicious or the network is compromised, users can always recover their assets by going to Ethereum as a court of higher appeal. This differs from some other roll-up solutions that cannot currently implement such enforcement mechanisms.

L2 platforms are more centralized than L1s, with key risks including: (1) Centralized sequencer/proposer risk - if core development teams leave, users need mechanisms to continue operation; (2) Exit mechanisms - users need ability to exit if the chain is going in unwanted directions; (3) Smart contract modification risk. Validation is achieved through optimistic setups (simpler) or zero-knowledge setups (better for privacy). Banks are deploying products like JP Morgan's JPMD on Base, providing openness of crypto with most Ethereum security plus identifiable centralization that can be mitigated. This recreates a situation comparable to existing operational risk frameworks where contracts define remediation responsibilities.

A rollup is a subset of L1 data defined by a state transition function, where users post data to L1 and rollup nodes execute based on that data. A sequencer handles batching, ordering, posting, and committing transactions to L1, making it central to censorship resistance and MEV extraction. Four sequencer architectures exist: centralized/non-shared, centralized/shared, decentralized/non-shared, and decentralized/shared. The industry is moving toward decentralized and shared architectures for censorship resistance, reduced MEV centralization, cross-rollup composability, and easier new rollup launches. Key design considerations include liveness, censorship resistance, data availability, faster than L1 block time, and good user experience. Alternative approaches include L1-sequenced rollups (limited by L1 block time), escape hatches for censorship (sacrifices economic benefits), and governance-controlled centralized sequencers (still centralized with governance delays). A staking-based selection mechanism enables permissionless participation through single leader election. For shared sequencers, lazy execution is necessary because executing rollup state machines directly would make the sequencer too heavy. The sequencer handles batching, ordering, and committing without executing rollup logic, treating transactions as opaque bytes. Rollup full nodes execute transactions lazily when needed, and the sequencer is agnostic to how state roots are committed (optimistic or ZK proofs).

Sequencers are centralized components that order transactions before they are posted to the data availability layer. They enable pre-confirmation, allowing users to receive updates within milliseconds (5ms latency) before full consensus occurs. The sequencer acts as a gatekeeper, ordering transactions and sending them to proposers who post them to Solana. While centralized, there are escape hatches—users can always interact directly with L1 if the sequencer censors them. The architecture can transition to decentralized sequencers as technology matures.