Proto-danksharding

Danksharding is the blunt instrument the network needed, a technical alley where rollups hand over their baggage so the main chain can breathe. Proto-danksharding is the first step down that alley. It is the specification that introduces blobs and blob-carrying transactions so layer two bundles can post large amounts of data cheaply and temporarily instead of forcing the base layer to hoard history forever. Blobs are short-lived binary stores. They hold up to 128 kilobytes each and they live for a few thousand epochs before being pruned, which keeps long-term node storage low. Blocks can carry multiple blobs so a single block can absorb far more rollup data than before. Blob-carrying transactions are built to carry a commitment to those blobs without forcing every full node to read the full content. The commitments use polynomial commitment schemes and zero-knowledge techniques so the execution layer can trust that the data exists while avoiding permanent bloat. There is a separate pool for these transactions so they do not compete directly with ordinary execution traffic. Fees for blob space use a market-driven mechanism inspired by prior fee upgrades so pricing reacts to demand for temporary data slots. The practical effect is simple to say and hard to overstate: rollups get cheaper to operate and the mainnet stops choking on calldata, which reduces congestion, lowers effective fees across the ecosystem and speeds up finality for users. The trade-offs are blunt as well. Off-chain validation and lighter on-chain storage increase the attack surface for less popular rollups and for any infrastructure provider that sequences or posts data. Smaller rollups with thin security budgets could be probed or exploited, and cross-layer trust models must be tightened. Mitigations exist in the toolbox. Data availability proofs, sampling, stronger incentive designs and robust fraud or validity proof systems can push risk back down to acceptable levels. Proto-danksharding is not the end state. It is scaffolding toward full danksharding, which aims to shard data more deeply and scale parallelism even further. For users and builders the takeaway is this. The main chain will stop insisting on forever storage for every rollup batch. That change makes layer two ecosystems cheaper and faster. It does not remove the need for careful security design at the rollup layer. The future will be a layered stack where ephemeral blob space on the base layer feeds many more, leaner rollups, and where the network trades history for throughput to serve a larger, stranger world of decentralized apps.