Ref Node Live| VIEW RUST CORE ON GITHUB →
INTRODUCTION
CHAIN
RESOURCES
INNOVATION
CONNECT
Infrastructure

The Node Stack

A protocol is only as strong as the nodes that run it. Sahyadri defines three node modes with different trade-offs in storage, verification cost, and trust. None of them trust any other node. Each mode verifies the protocol rules that its role requires, and none requires permission to run.

01 · Node Modes

Three Trade-offs, Zero Trust

The three node modes differ in how much of the chain they store and which proofs they accept. They do not differ in whether they trust peers — none of them do. A full node re-executes every transition. A pruned node verifies every block in its window. A light client verifies every proof it accepts.

Full Node

Holds
Full state, from genesis
Verifies
Every block, every transition
Purpose
Reference implementation. Re-runs the transition function on every block.
trust = 0

Pruned Node

Holds
Recent state, older state pruned
Verifies
Every block within its window
Purpose
Runs on smaller hardware. Verifies recent history independently.
trust = 0 | window

Light Client

Holds
No chain state
Verifies
Proofs against known roots
Purpose
Verifies state and inclusion using recursive proofs. Runs on a phone.
trust = 0 | proof
02 · Mining Nodes

Miners are not Validators

A validator, in the PoS sense, is a party selected by stake to produce blocks. Sahyadri has no such role. The parties that produce blocks are miners — they produce blocks by finding a valid PoW nonce, not by being selected. There is no validator set, no election, no stake, no slashing.

produce(block) ⟺ ∃ nonce : pow(nonce) ≤ target

The condition for producing a block is a mathematical inequality — nothing else. No registration, no stake, no approval, no identity. A miner is anyone whose hardware finds a nonce that satisfies the inequality. When they find it, they broadcast the block, and every node in the network independently verifies that the inequality holds.

The Mining Condition

Sahyadri uses the term "miner" throughout — never "validator." The two are different roles with different security assumptions. Confusing them is how projects end up describing PoS mechanisms in a PoW system. The protocol has mining nodes. That is all.

03 · Storage

Two Storage Paths

Different node roles have different storage requirements. A pruned node keeps only the recent state needed to validate current blocks. A full node keeps the current state and the block history needed to re-verify. An archive node keeps everything. The protocol defines what the state is; the node implementation decides how to store it efficiently.

I

Current State

The current account balances and Crest objects. This is what pruned, full, and archive nodes all keep — because it is what validation requires.

Required by every node mode
II

Historical Blocks

The full sequence of blocks and their headers. Required to independently verify the chain from genesis or from any checkpoint.

Required by full and archive nodes
III

Historical State Snapshots

The state at each historical block. Required to answer queries about past balances, DIDs, and credentials without replaying.

Required by archive nodes only

The specific storage engine and storage format are implementation details, not protocol constants. Different implementations may use different storage backends — the protocol defines what must be verifiable, not how it must be stored.

04 · Network Layer

A Peer-to-Peer Topology

Nodes communicate directly with each other over a P2P protocol. There is no central hub, no relay, no privileged node. A new node connects to any peer it can reach, syncs up, and begins participating. The graph formed by peer connections determines how quickly information propagates and how resilient the network is to partition.

Tpropagate = O( log N )

The propagation time from any node to any other is logarithmic in the network size. Doubling the network adds roughly one more hop, not one more unit of latency. This is why a decentralized network can achieve near-instant broadcast without a central authority — the topology itself provides the efficiency.

Each node maintains multiple peer connections so that no single peer is a critical bridge. When a peer disconnects, the node reconnects to another. When a partition occurs, both sides continue producing blocks — and when the partition heals, both DAGs merge and the canonical order reconverges. This is a property of the DAG, not a feature that nodes must negotiate.

05 · Participation

Three Ways to Contribute

Running infrastructure on Sahyadri requires no permission. There is no registration form, no identity verification, no acceptance process. You download the software, run it, and the network accepts you because the software behaves correctly.

Run a Node

Download the binary, run it, and start verifying. Every additional node increases the network's partition tolerance and decreases the risk of any single node being a point of failure.

Requires: Commodity hardware
security(N) ∝ N

Mine

Point a miner at a running node. The miner searches for valid PoW nonces and submits blocks. No pool required, no registration required.

Requires: CPU + 16 MB RAM/thread
reward(h) = R₀ · 2^(-⌊h/H⌋)

Build Tooling

Wallets, explorers, indexers, SDKs. Every tool that makes the protocol easier to use reduces friction for the next participant.

Requires: Developer time
usability ∝ 1/friction
Infrastructure is not what a foundation owns. It is what anyone can run. A node is not a validator — it is a witness, and its word is the protocol.
— Sahyadri