Fountain Codes: a way to reduce blockchain storage costs

The 246/10 system employs Galois Field (GF(2^8)) and utilizes 10-byte code words, ensuring each block is padded to a multiple of 10 bytes for consistency and integrity. In this system, archival nodes serve as primary data holders, providing shards numbered 0-9 directly. Additionally, each sharded node offers a single shard from a range of 10-255 for each block. These sharded nodes are organized into 246 distinct anonymity sets, with each set managing significantly reduced blockchain storage of 70GB compared to the standard 700GB. This structure results in a weekly data increase of approximately 200MB.

In practice, a user connects to peers that hold twelve different shards, such as [15, 25, 28, 36, 39, 130, 137, 138, 162, 176, 194, 222]. To verify data integrity, a user might request ten of these peers to provide their respective shards for a specific block, e.g., block 420,004. Each shard typically comprises 87,166 bytes after appropriate padding. Should the combined shards fail to produce the correct hash, indicating potential malice or errors, the user then requests shards from the remaining two peers to attempt error correction. Successful error correction allows the identification and exclusion of the problematic peer. If unsuccessful, the user may either expand their network by including more peers or resort to downloading the block from a reliable archival node to pinpoint the errant shards.

This approach not only simplifies the process by requiring connections to fewer peers—about 10 or 12 instead of 80 or more—but also imposes a higher storage demand. While it saves approximately 90% of storage space, it does not achieve the up to 99% savings seen in other specifications. Moreover, this method ensures there is no data inflation when combining differently sized blocks, maintaining strict data management protocols.