Chunked Uploads¶

Introduction¶

Every Bulletin Chain storage extrinsic carries a hard per-transaction byte limit. A payload smaller than the limit goes in a single transaction. A payload larger than the limit is chunked: the SDK splits the bytes into chunks, uploads each chunk in its own transaction, and writes a small DAG-PB manifest describing how the chunks reassemble. The CID returned for a chunked upload is the CID of the manifest, not of any individual chunk; readers fetching that CID get the manifest, follow it to the chunk CIDs, and reassemble the original payload.

For most Products, this is transparent — app.bulletin.upload(bytes) accepts a payload of any size up to the SDK's supported maximum, and the chunking pipeline runs underneath. This page documents what the pipeline does and what trade-offs it surfaces.

What Happens Internally¶

The SDK's chunking pipeline runs roughly this sequence:

1. Split the payload into fixed-size chunks (the chunk size is the SDK default, sized to fit comfortably under the per-transaction byte limit).
2. Upload each chunk in its own extrinsic, accumulating CIDs as the chain returns them.
3. Build a DAG-PB manifest a small object that lists the chunk CIDs in order, encoded in the DAG-PB format that IPFS-style readers recognize.
4. Upload the manifest itself as a final small transaction.
5. Return the manifest CID to the Product.

Readers retrieving the manifest CID get back the manifest first, then issue parallel fetches for each chunk CID and reassemble. Because both the chunks and the manifest are content-addressed, readers can verify each piece independently.

What This Costs¶

Chunked uploads cost more than single-transaction uploads because each chunk consumes one transaction from the authorization quota. A Product uploading large payloads on behalf of many users should budget transaction quota accordingly — running out of transactions even with plenty of bytes remaining is a realistic failure mode.

The manifest itself counts as one additional transaction on top of the chunk count.

What a Product Sees¶

A Product calling app.bulletin.upload(bytes) sees the same shape regardless of whether chunking ran underneath:

• The call returns a CID.
• The call decremented the account's authorization quota by the appropriate amount (1 + chunk count for a chunked upload, 1 for a single-transaction upload).
• A subsequent app.bulletin.fetch(cid) retrieves the original payload.

The chunking is an implementation detail of the upload pipeline, not part of the surface a Product needs to manage. For Products that want explicit control (custom chunk size, parallelism limits, progress reporting on large uploads), the Store Data on Chain guide covers the lower-level BulletinClient and the chunk-level controls it exposes.

Where to Go Next¶