Storage reliability · 2026 · Solo — storage-systems research, finite-field engine, interaction design, test-first build, verification

Parity Deck

A browser-native recovery lab that separates erasure reconstruction from integrity detection and fails closed on insufficient trusted shards.

The problem

Erasure coding is often explained as a capacity ratio rather than an operational recovery decision. Parity Deck exposes four data and two parity cartridges, lets an engineer mark shards missing or corrupt, quarantines CRC failures, and shows the exact survivors and inverse matrix used to restore one fixed archive record—or the precise reason recovery cannot proceed.

Architecture

Key decisions

01

Separate corruption from absence

Missing shards reduce availability; corrupt shards are present but excluded by trusted CRC metadata. The recovery receipt shows both states rather than conflating them.

02

Show the actual decoder

The interface exposes selected survivor IDs and the 4×4 inverse matrix so recovery is inspectable rather than a decorative simulation.

03

Bound the claim

Parity Deck uses one short payload and a fixed systematic 4+2 Vandermonde-derived layout. It is not Ceph-compatible and does not claim protection from unknown silent corruption without trusted checksums. Seven tests and production browser probes passed; current source is SHA 5f7fe0493486cd5391d34312a0a1f7a5ac468b2f.

Metrics

7
engine tests passing
6
interactive shards
4
trusted survivors required