Data Availability
Data availability (DA) is the guarantee that the transaction data behind a rollup's state is published somewhere anyone can read it — so that anyone can independently reconstruct and verify the rollup's state. The RDK defines three DA backends; today only native is implemented.
| Backend | What it is | Status |
|---|---|---|
native | On-chain blob storage within QoreChain itself | Implemented in the module; no public way to post a blob yet (see below) |
celestia | Data availability via IBC to Celestia, a dedicated modular DA layer | Not implemented |
both | Native and Celestia together, for redundancy | Behaves exactly like native today |
No rollup has been created on either network yet (checked 27 October 2026), and the DA backends are only partly implemented. The status of each backend is stated below; validate on the qorechain-diana testnet before relying on any of them.
Native DA (on-chain blob storage)
Native DA is the on-chain blob store inside x/rdk: each blob is keyed by rollup ID and blob index, carries a SHAKE-256 commitment of its bytes, and is pruned after the retention window while its commitment is kept.
From release v3.2.8 (testnet since 12 November 2026; mainnet has not applied it) a rollup can post blobs with the message MsgPostRollupData, and RDK 0.9.0 supports it. On releases before v3.2.8, including mainnet today (v3.2.7), the module defines blob storage, retention and pruning, but there is no transaction message for posting a blob (the module's messages were checked in the chain source at v3.2.5, and neither the v3.2.6 nor the v3.2.7 release notes add one); the rest of this note describes those releases. The chain accepts a non-empty data_hash on a native or both rollup only when a matching blob is stored on chain, and nothing can store one. So a settlement batch on such a rollup must leave data_hash empty: a batch that names a data hash is refused as data unavailable, after the transaction fee has been paid.
The RDK follows this from 0.6.1 (and so do 0.6.2 to 0.8.0): SubmitBatchInput.dataHash is optional (omitted means empty), all six starter templates leave it empty, dataHashError (data_hash_error, DataHashError, Rules.dataHashError) exposes the rule in all five languages, and RdkTxClient.submitBatch refuses such a batch before broadcasting when it can read the rollup (a hybrid client with a REST endpoint). Earlier template releases put a data hash on their batches, which the chain refused.
Key concepts:
- Blobs. A blob is associated with a rollup ID and a blob index.
- Commitments. Each blob carries a 32-byte SHAKE-256 commitment of its data, so a blob can be verified against what was committed, without trusting the storer. (The TypeScript helper
buildDaBlob()hashed with SHA-256 in 0.6.2 and earlier, so itsdataHashdid not match the chain's commitment; from 0.7.0 its commitment is the chain's SHAKE-256. Since no transaction posts a blob, you still should not send a data hash in a batch on anativeorbothrollup.) - Namespaces. Blobs can carry a rollup-specific namespace, keeping each rollup's data logically separated within shared storage.
- Retention and pruning. Native blobs are retained for
blob_retention_blocks(432,000 blocks — about 16 days at mainnet's current ≈3.1 s block time; the figure moves with block time) and then pruned to keep on-chain storage sustainable. On chain v3.2.8 the default retention for posted data is 648,000 blocks (maximum 25,920,000), and it is never shorter than the challenge window. After pruning, the raw blob data is removed while commitment metadata is retained, so the historical commitment remains verifiable even though the bytes are no longer stored on-chain.
The maximum blob size and retention window are governed by the live module parameters. Query them before you design around any specific limit:
qorechaind query rdk config
Blob metadata (not the bytes) is readable over REST at /qorechain/rdk/v1/blob/{rollup_id}/{blob_index}, which both networks serve; it includes expires_at_height.
Celestia DA (IBC to Celestia)
Celestia DA is not implemented. A rollup whose DA backend is celestia cannot post data (the module returns an error). The planned design publishes data availability through IBC to Celestia, a dedicated modular DA network, while still anchoring settlement on QoreChain; no IBC channel is open on mainnet, and the one open channel on the testnet (checked 27 October 2026) is a token-transfer (ics20-1) channel, not a data-availability channel (see Bridge Architecture). Use native.
Both (redundancy)
The both backend is intended to write to native and Celestia together. Today both behaves exactly like native: only the native copy is stored, and the Celestia half is skipped with a logged warning. Do not treat it as two independent copies.
Choosing a backend
| If you want... | Choose |
|---|---|
| The simplest, fully on-chain option inheriting QoreChain security | native |
| To offload DA to a dedicated modular layer | celestia — not implemented |
| Redundancy across two stores | both — same as native today |
For how DA fits into the broader settlement picture, see Rollups Overview. For the lower-level module reference, see the Rollup Development Kit page.