Signal, Matrix, and the Protocol Wars

01 When the protocol is the politics

Every chat app you've ever used rests on a protocol — a rulebook that defines how messages are formed, routed, encrypted, and stored. Most people never think about this layer. But if you care about who controls your conversations, it matters enormously, because what a messaging protocol is shapes what a messaging app can and can't do to you.

Two protocols have come to define the privacy-conscious end of the modern chat landscape. Signal Protocol, developed by Open Whisper Systems and now stewarded by the Signal Foundation, is the tightly engineered heart of the Signal app — and, quietly, the encryption layer inside WhatsApp, Facebook Messenger's secret conversations, and Google Messages. Matrix, created by Amdocs engineers Matthew Hodgson and Amandine Le Pape and launched in 2014, is an open federated standard with no single owner and a reference app called Element. They share a commitment to private, secure messaging. Almost everything else about them is different — and that difference is, at its core, political.

02 Signal Protocol: the locked safe that everyone borrows

Signal Protocol is not something you run; it is something you use, probably without knowing it. The protocol's core innovation is the Double Ratchet Algorithm, developed by Trevor Perrin and Moxie Marlinspike around 2013. Pair it with the X3DH (Extended Triple Diffie-Hellman) key agreement mechanism and you get forward secrecy — the property that means even if an attacker records your ciphertext today and steals your keys tomorrow, they cannot decrypt yesterday's messages. Each message generates fresh key material. This is cryptographically elegant, and it is why almost every serious messaging app has either adopted Signal Protocol outright or reverse-engineered its best ideas.

The Signal app itself is deliberately minimal. Its server infrastructure is centralized — run by the Signal Foundation, a non-profit — and by design reveals almost nothing. The app's Sealed Sender feature obscures even who is messaging whom. Its metadata footprint is famously small: when the US government subpoenaed Signal, it could supply only the account creation date and the date of last connection. That is not a PR talking point; that is a documented court outcome. The Foundation has made "collect less" the architectural philosophy, and the architecture enforces it rather than promising it.

But Signal Protocol's strength is also its constraint. The Signal app is a walled garden, intentionally so. You cannot run your own Signal server and connect to Signal users. There is no federation, no interoperability at the protocol level between Signal and anything else. Moxie Marlinspike argued this position explicitly: federation ossifies protocols, makes iteration slow, and lets the perfect be the enemy of the good. If Signal had been federated from the start, he contended, it would never have adopted Forward Secrecy or sealed sender without dragging every third-party server operator along first.

That argument is coherent. It is also a concentration of trust. You are trusting the Signal Foundation not just with good intentions today but with good governance indefinitely — that it never faces regulatory pressure it cannot resist, never suffers a leadership failure, never changes its model. For now, those are reasonable bets. They are still bets.

The French government migrated its internal communications to a Matrix deployment.

03 Matrix: the open mesh

Matrix approaches the problem from the other direction. Rather than build a trustworthy institution and minimize what it knows, Matrix distributes trust so thoroughly that no single institution can be the chokepoint. It is a protocol for federated, persistent, replicated conversation — closer in philosophy to email or IRC than to a conventional messaging app.

In the Matrix model, anyone can run a homeserver. Users on one homeserver can talk to users on any other. Conversations are not stored in one place; each participating homeserver holds a copy of the room's event graph, a cryptographic structure that makes the history tamper-evident. This is robust against takedowns — no single server going dark can destroy a room that other servers have joined. It also means Matrix rooms have a fundamentally different privacy character from Signal conversations: the data exists on every homeserver whose users participate, and the operators of those servers can see unencrypted metadata and, if encryption isn't used, content.

Matrix does support end-to-end encryption, using a system called Megolm (a group ratchet that builds on Olm, Matrix's own implementation of the Double Ratchet concept). In encrypted rooms on Element or other clients, message content is protected from homeserver operators. But the federation architecture itself generates metadata — who joined what room when, from which server — that is inherently more visible than Signal's sealed-sender model. Choosing Matrix means choosing resilience and openness over the tight metadata minimization that Signal enforces by construction.

Where Matrix genuinely shines is in scale, complexity, and institutional use. The French government migrated its internal communications to a Matrix deployment. The German armed forces, the Bundeswehr, runs a Matrix-based platform. These organizations need to own their infrastructure, federate across departments, and not depend on an American non-profit's continued goodwill. Matrix is the right answer for that requirement in a way Signal structurally cannot be.

04 Why it matters to you

The "protocol wars" framing is slightly dramatic — Signal and Matrix are not really competing for the same users in the same markets. But the tension between them maps onto a genuine disagreement about what security and privacy mean in practice, and that disagreement has consequences.

If your threat model is mass surveillance, a government subpoena to a service provider, or a data breach at a centralized company, Signal's architecture is almost uniquely well-suited. The data that doesn't exist can't be handed over and can't be breached. Signal's minimalism is protective minimalism.

If your threat model is platform risk — a single company or foundation failing, being acquired, being regulated into backdoors — Matrix's distribution is the answer. No one can turn off Matrix. Individual servers can be seized or shut down; the protocol and the mesh cannot be. For communities, for institutions, for people who need to own their communications infrastructure, that matters more than sealed sender.

There is a third consideration, and it is the EU's Digital Markets Act. Designated gatekeepers — the largest platforms — are now required to offer interoperability with third-party messengers. How that plays out technically is still being negotiated, but both Matrix and Signal Protocol are being cited in those conversations. Matrix's federation model is a natural fit for DMA-style interoperability. Signal has resisted, arguing that mandatory interoperability would fatally compromise its security guarantees — and that argument is not without merit, because bridging to a less-secure protocol weakens the chain.

05 Two rights, one hard choice

The honest summary is this: Signal Protocol is more private, by measurable architectural criteria. Matrix is more open, by design. Signal asks you to trust the Foundation; Matrix asks you to trust the mesh — and to accept that a mesh is inherently more legible than a locked safe.

Neither is wrong. The choice depends on what you are protecting yourself against, and the federated vs. centralized trade-off is one every serious user eventually faces. What matters is that both visions exist, that both are open-source and auditable, and that both have pushed the mainstream — WhatsApp, iMessage, Google Messages — toward meaningfully better encryption than existed a decade ago.

The protocol wars are not won. They are, in a real sense, the productive argument that keeps both sides honest.

◆ A short timeline

1. 2013Double Ratchet Algorithm developed; foundation of Signal Protocol
2. 2014Matrix protocol publicly launched by Hodgson and Le Pape
3. 2016WhatsApp completes Signal Protocol integration for all message types
4. 2018Signal Foundation established as non-profit to steward the app
5. 2022–ongoingEU Digital Markets Act drives interoperability debate

◆ People & organisations