“Avail: A Scalable Data Availability Layer” is the 2022 reference document originally produced by Polygon Technology (now published by the independent Avail Project) describing Avail — a purpose-built modular data availability blockchain. Avail’s sole function is to order transaction data blocks and guarantee their availability using a combination of KZG polynomial commitments, 2D Reed-Solomon erasure coding, and Data Availability Sampling (DAS) for light clients.

Avail enables rollups, validiums, and sovereign chains to post transaction data to a secure DA layer without inheriting the cost of Ethereum L1 storage, while providing cryptographically verifiable availability guarantees.

> Reference Document: Available at GitHub/availproject.

---

Publication and Context

Avail was originally designed as a component of Polygon 2.0 — the “Polygon DA” layer — but was spun out as a standalone project after Polygon pivoted its modular strategy. As an independent project, Avail competes with Celestia and EigenDA as a modular DA provider.

The modular blockchain thesis (Celestia, 2019): separating consensus, execution, and data availability into different layers allows each layer to specialize. Avail specializes entirely in DA, providing this service to any execution layer, including EVM rollups, Substrate chains, and sovereign application chains.

---

Data Availability Problem

When a block producer publishes a block, it must be verifiable that:

1. The block data is actually available (not withheld)
2. The data is correctly encoded (not corrupted or malformed)

Without DA guarantees, a dishonest block producer could publish a block header with a valid commitment but withhold the underlying data — preventing observers from detecting fraud or generating fraud proofs.

Avail’s solution: Construct blocks so that light clients can verify availability by sampling random small chunks — without downloading the entire block.

---

2D Reed-Solomon Erasure Coding

Avail uses 2-dimensional Reed-Solomon erasure coding rather than 1D:

1. Block data is arranged into a k×k matrix
2. Each row is extended from k to 2k symbols using Reed-Solomon (1D erasure coding)
3. Each column is also extended from k to 2k symbols
4. The result is a 2k×2k matrix where the original data is recoverable from any quadrant of sufficient size

Why 2D? With 1D erasure coding, a malicious block producer can withhold a contiguous segment that prevents a specific fraud proof. With 2D coding, any (k+1)×(k+1) submatrix is sufficient to reconstruct the full block, and random sampling provides very high probability of detecting withholding.

---

KZG Commitments

For each row and column of the extended 2D matrix, Avail computes a KZG polynomial commitment:

• Row i is treated as polynomial $p_i(x)$; the commitment $C_i = [p_i(tau)]_1$
• Column j is treated as polynomial $q_j(x)$; the commitment $D_j = [q_j(tau)]_1$
• Both row and column commitments are published in the block header

Light client verification:

A light client samples a random cell (i, j) and receives:

• The cell value
• A KZG proof that the value is consistent with both the row i commitment and the column j commitment (checking both ensures the cell is genuinely encoded in the 2D matrix)

A light client that samples 40 random cells achieves >99.9999% probability of detecting a block where >50% of cells are withheld.

---

Nominated Proof of Stake

Avail’s validator consensus uses Nominated Proof of Stake (NPoS) — the same model used by Polkadot — via a BABE + GRANDPA consensus combination:

• BABE (Blind Assignment for Blockchain Extension): VRF-based leader selection for block production; provides probabilistic finality
• GRANDPA (GHOST-based Recursive Ancestor Deriving Prefix Agreement): Byzantine Agreement protocol for deterministic finality over chains of BABE blocks

Nominators stake AVAIL tokens to back validators. Validators who misbehave (equivocate or withhold data) are slashed, with a portion going to their nominators.

---

Application-Specific Namespacing

Like Celestia, Avail supports data namespacing: each block can contain data blobs tagged with application-specific namespace identifiers. A rollup using Avail only needs to download blobs with its own namespace — not the full Avail block.

This allows multiple rollups to share a single Avail block without paying for or processing each other’s data.

---

Reality Check

Avail represents a well-engineered modular DA layer. The 2D erasure coding + KZG approach is technically superior to simpler DA designs for light clients. However:

• Market competition: Avail competes with Celestia (first-mover, established ecosystem) and EigenDA (Ethereum-secured). Capturing market share against both established competitors is challenging.
• Validator set bootstrapping: Avail mainnet launched in 2024 with a small initial validator set; achieving meaningful stake distribution takes time.
• Light client deployment: The DAS advantage only materializes when rollup light clients actually use DAS. Most rollup nodes run full nodes that download all data regardless.

---

Legacy

Avail’s 2D Reed-Solomon + KZG approach is the most sophisticated light-client DA sampling design in production (combining improvements from both the Celestia and Ethereum DA research lines). The project demonstrated that Polygon’s modular research translated to a standalone competitive DA layer. Avail’s NPoS model provides a well-studied staking design with known properties from Polkadot.

---

Related Terms

• Data Availability
• Celestia Whitepaper
• EigenDA Whitepaper
• KZG Commitment
• Erasure Coding

---

Research

• Paul, A., et al. (2022). Data Availability — Reference Document. Avail Project.

— Primary technical reference. Section 2 defines the 2D RS coding; Section 3 covers KZG row/column commitments; Section 4 describes light client sampling.

• Al-Bassam, M. (2019). LazyLedger: A Distributed Data Availability Ledger With Client-Side Smart Contracts. arXiv:1905.09274.

— Celestia predecessor paper; the original formalization of namespaced DA sampling that influenced Avail’s design.

• Dankrad, F. (2020). Data Availability and Erasure Coding. Ethereum Research.

— EF Research post on 2D erasure coding for Ethereum’s danksharding; closely related to Avail’s 2D RS design.