Back to: Certified Decentralization Analyst
Software is the most overlooked pillar — and the most dangerous one. It asks: how many independent versions of the node software exist, and are they actually in use? A chain can be perfectly distributed in hardware and stake, and still have a single point of failure in code.
Why client diversity matters
If 100% of nodes run the *same* software, then one bug in that software is a bug in the whole network: – A consensus-breaking bug in one buggy client takes down the *entire* chain — a self-inflicted 51%-style outage with no attacker. – One maintainer or company controls upgrades of all clients = governance by codebase.
What we actually measure
– Number of independent clients (e.g., Ethereum consensus: Prysm, Lighthouse, Teku, Nimbus…). – Usage share — the real risk number. Diverse *options* with 95% running one client is still a 1-client network. – Shared code/stack — “N different clients” is weaker if they share libraries, specs, or a foundation treasury.
Reading the scores
| Score | Reading | |
|---|---|---|
| Ethereum | 88/100 | Best-in-class client diversity — several healthy consensus clients |
| Bitcoin | 70/100 | Bitcoin Core dominates ~95%; Knots & btcd are minorities |
| Near | 55/100 | A single primary client codebase |
| Polkadot | 68/100 | Strong client ecosystem, still heavily weighted to one codebase |
The analyst's checklist
1. List the *implementations*, not the marketing page. 2. Find the real usage split (telemetry, block production share) — block- producing share is the truth. 3. Check whether the “independent” clients actually share code.
“We have 3 clients!” is marketing. “Most blocks are made by client X” is a
> fact. Use the fact.
Next lesson: a hands-on lab — scoring a chain in the terminal.