Introduction

This document defines the deterministic subnetwork assignment algorithm used in the Nomos Data Availability (DA) network. The DA network is composed of multiple subnetworks, and each participating node must be assigned to one or more subnetworks to ensure data redundancy, availability, and sampling coverage.

The objective of the assignment algorithm is to:

Maintain stability across session, minimizing changes to existing node-to-subnetwork relationships.
Efficiently assign new nodes and remove inactive ones.
Guarantee deterministic results across all honest participants, enabling independent computation of subnetwork assignments without coordination.
Preserve balanced subnetwork sizes and ensure each subnetwork meets a target replication factor.

This specification is used in conjunction with the Service Declaration Protocol which provides the list of active nodes eligible to participate in the DA network.

Overview

This algorithm distributes DA network participants across various subnetworks. Participants are selected exclusively from the Service Declaration Protocol listing for DA, meaning only nodes included in that listing qualify for distribution.

At each session, the algorithm takes as input:

The list of currently declared active nodes (ProviderIds) from the SDP.
The previous subnetwork assignments (from the last epoch).
A desired replication factor (minimum number of nodes per subnetwork).
A deterministic source of randomness (Blake2b-based PRNG seeded with session randomness).

The goal is to produce a new assignment that:

Retains active nodes in their previous subnetwork(s) where possible.
Removes any nodes that are no longer active.