Walrus 2D Encoding: Primary Slivers + Secondary Repair Done Right
Two-dimensional encoding separates data into primary and secondary layers, but implementation matters enormously. Poor separation creates false efficiencies. Walrus gets this right through careful structural choices.
Primary slivers are fragments created directly from original blob data with minimal transformation. They are smaller than secondary slivers and fast to serve. Clients preferentially request primaries because they offer lowest-latency recovery. The system is designed so primary slivers are distributed across validators most likely to be responsive and well-connected.
Secondary repair slivers are mathematically derived from primaries, providing redundancy for Byzantine resilience.
They are larger and more computationally expensive to compute but necessary for guarantees. Secondary slivers are positioned strategically to cover gaps left by primary unavailability.
The elegance is structural separation that aligns with actual network conditions. Fast validators get primary responsibility. Slower validators hold secondaries. Recovery naturally gravitates toward fast path through primaries, but secondaries are available if needed.
Most schemes blur this distinction. @Walrus 🦭/acc makes it explicit. Every sliver's role is mathematically defined. Primary fragments are smaller, faster, and numerically sufficient when available. Secondaries provide complete coverage when primaries are unavailable.
This clarity enables optimization. The protocol routes around congestion automatically by preferring primaries over secondaries. Clients always get optimal recovery path based on actual network state.
