EntropyFS: Entropy-Native Configurational Storage as a Filesystem Substrate - Broad Prior-Art Technical Disclosure and Research Architecture
Riaan De Beer · Zenodo (CERN European Organization for Nuclear Research) · 2026
Contemporary filesystems generally expose byte-addressed objects and persist those objects through some combination of blocks, extents, compression, deduplication, sparse allocation, copy-on-write sharing, and metadata indirection. These techniques can substantially reduce physical storage relative to logical size, but the persisted representation remains conceptually close to the materialized byte sequence: bytes are compressed, blocks are shared, ranges are omitted, or differences are encoded. This disclosure explores a broader storage abstraction called entropy-native configurational storage, instantiated here as the conceptual filesystem EntropyFS. The central proposal is that application-visible bytes should be treated as a materialized interface, while the persistent storage medium is a bounded, deterministic representation graph consisting of entropy-coded states, combinatorial ranks, immutable references, deterministic generators, transforms, shared bases, and exact residuals. An extent need not exist verbatim on the underlying SSD, NVMe device, HDD, or RAM pool provided that an exact, bounded and independently verifiable materialization procedure reconstructs it. The proposal does not imply that information can be stored below its information-theoretic description length, nor that a small random seed can encode an arbitrary much larger incompressible object. A descriptor containing \(k\) independent bits can distinguish at most \(2^k\) states unless additional persisted or shared information contributes to the reconstruction. Shannon source coding remains a fundamental boundary, and the shortest universal description of an arbitrary string is related to Kolmogorov complexity rather than to the expanded size produced by a deterministic generator. The proposed novelty of investigation is therefore not “infinite compression.” It is the systematic elevation of mathematical configuration and residual state to filesystem storage primitives. The filesystem searches a bounded representation space and persists the smallest useful exact description it can discover under explicit constraints on storage size, read cost, write cost, dependency depth, and computational complexity. Several established fields supply important foundations: Shannon coding, asymmetric numeral systems, enumerative source coding, succinct data structures, dictionary and grammar compression, content-addressed storage, copy-on-write snapshots, deduplication, compressed filesystems, compressed memory, delta transfer, and deterministic generative descriptions. EntropyFS is positioned as a synthesis and extension of these principles rather than as a rejection of them. A second contribution explored here is the use of deterministic residual-structure observation to guide representation search over time. The existing DSFB implementation exposes residuals, aggregate residuals, state, and per-channel trust explicitly, while related DSFB work treats drift, slew, persistence, and other residual motifs as analyzable structure. In EntropyFS this machinery is proposed strictly as an optimization observer: it may decide which representation candidates deserve evaluation, but it is never allowed to determine decoded data. Correctness remains purely deterministic and byte-exact. The result is a broad architecture spanning persistent filesystems, userspace block devices, and potentially RAM: persist irreducible description state; regenerate deterministic structure; represent change as residual structure; and retain raw bytes whenever nothing better exists. Keywords: entropy-native storage; configurational storage; filesystems; source coding; enumerative coding; asymmetric numeral systems; content-addressed storage; copy-on-write; deduplication; residual analysis; compressed memory; deterministic materialization; representation graphs; prior art.