Blockchain Direction

Offline-first multi-channel propagation on nexus, chton, and Tagma

Author
Affiliation

SSCCS Initiative

SSCCS Initiative

Abstract

This is a vision draft. Blockchains inherit a fundamental availability problem: when the wired path to the consensus network is cut, state sync, broadcast, and finality all stop, and asset access is blocked even though keys stay local. The nexus record layer after the L2 restructure already carries the primitives of a content-addressed, append-only ledger: full-injective CoordId<20> ids, deterministic rebuild_cache replay, and partition-based channel semantics. This document sketches a staged development direction toward an offline-first, multi-channel propagating blockchain: a deterministic state root first, then DAG reconciliation over content-unique facts, then a propagation layer, and only then optional ordering. The propagation and aether-linked transport parts are conjecture; the state root and reconciliation steps are concrete and useful regardless of the blockchain goal.

The Availability Problem

A blockchain’s usable state depends on reaching the consensus network. Three operations stop when the wired path is cut: synchronizing state, broadcasting transactions, and verifying finality. Keys are local, but the ledger is not, so an owner cannot prove or move value without a network path.

This is not a ledger problem, it is a propagation and ordering problem. Existing partial answers exist: satellite broadcast, mesh relay, delay-tolerant networking. They share a common shape: detached nodes accumulate records locally and reconcile when a path appears. What they lack is a content-addressed, conflict-defined record structure that makes reconciliation cheap and deterministic.

What nexus Already Provides

The nexus record layer, after the L2 restructure, is already shaped like a ledger at the data level.

Facts are append-only and content-unique. A FactRecord id is a full-injective CoordId<20> encoding of the content hash, so two distinct contents never share an id and a record is verifiable by hash without any network.

State is replayable. rebuild_cache reconstructs the in-memory record maps and structural index deterministically from IO. A replica that receives the same set of records rebuilds the same BoardState. This is the seed of offline state reconstruction.

Records are content-addressed by construction. Blobs are keyed by SHA-256, records by the injective id. Deduplication, verification, and merge are defined by the content, not by arrival order.

Channels exist as partitions. scan_partition already gives per-channel data flows, the seed of multi-channel semantics.

The FIH lifecycle maps to ledger patterns. The Blackboard Decentralization note formalizes this: a Fact is a state flag, submit_hint is an event signal, the from_facts grounding rule is a threshold trigger, and the intent lifecycle is a commit-reveal overlay. The medium is not yet a ledger, but the operations already are.

The Gap

The record layer has data-level ledger properties but no ledger protocol.

No consensus or global ordering. FihStorage is a single-owner execution unit with no internal concurrency. Instances are independent replicas, not participants in an agreement protocol.

No adversary model. The current design assumes a trusted single boundary. A propagating blockchain requires a Byzantine or at least a crash-tolerant threat model.

Finality, intent expiration, and spam resistance are open. These are listed as open questions in the Blackboard Decentralization note: confirmation depth for facts, unbounded accumulation of unclaimed intents, and unrestricted submit_fact.

The Vision: Offline-First Multi-Channel Propagation

The direction is a propagation model, not a wired-replacement chain. Nodes operate detached, accumulate content-addressed facts, and reconcile on reconnect. Because record ids are content-unique, reconciliation is conflict-defined: the same record is deduplicated, distinct records are merged, and only ordering needs a resolution rule.

Multi-channel propagation means the medium is not a single ordered stream. Partitions map to channels, and a future transport layer moves records across them. The aether-linked transport is beyond the verified scope of this document; the propagation part is conjecture until a concrete relay model exists.

Figure 1: Staged direction: detached accumulation, content-addressed merge, DAG reconciliation, then optional ordering.

A Staged Direction

Phase 1: deterministic state root. Hash-chain record ids and content hashes into a state root. rebuild_cache becomes replay-to-root verification: any replica proves its state by rebuilding and comparing roots. This is concrete and useful without any blockchain goal.

Phase 2: DAG reconciliation. Detached nodes exchange record sets. Content addressing makes merge conflict-free. Ordering becomes a DAG with an explicit resolution rule for concurrent intents.

Phase 3: propagation. A transport layer over channels with relay incentives and bandwidth bounds. Partitions define channels; the relay economy determines whether propagation is sustainable.

Phase 4: optional ordering. A consensus protocol of the BFT class, or proof-based ordering, only where finality is required. Offline modes degrade to DAG semantics without a total order.

Key Insights

Content addressing is the offline enabler. A record verifies by hash without the network; availability of verification does not depend on availability of the network.

Deterministic replay is the state-reconstruction primitive. A partial replica rebuilds the same state from the records it received, which makes reconciliation a data problem instead of a protocol problem.

Ordering is the hard problem, not storage. The nexus record layer already stores append-only content-unique state. The chain is one answer to ordering; DAG plus resolution rules is another.

The L2 restructure left the record layer ledger-shaped. Append-only facts, injective ids, replayable state, and channel partitions are the data foundation. The missing parts are protocol, not storage.

Future Work

  • Implement the deterministic state root and replay-to-root verification.
  • Define DAG merge semantics over content-addressed facts.
  • Map partitions to channels and prototype a propagation layer.
  • Resolve the decentralization open questions: finality depth, intent expiration, and spam resistance.

References