Federations Are Not Decentralization
There is an idea in Bitcoin so dishonest that I have trouble believing serious people defend it, yet they do. The idea is that a federation — a group of named or unnamed parties holding multisig keys — counts as a "decentralized" sidechain.
Paul Sztorc has been pointing out the absurdity of this for years. In his "Derangements" essay he calls federations "right next to the worst possible idea: giving all of your coins to someone else." That is not hyperbole. It is a precise structural claim. Let me explain why it is correct, and why the federated model has become one of the most damaging detours in Bitcoin's history.
What a federation actually is
A federation is a group of N parties holding a multisig wallet. To move funds out, M of N must sign. The marketing claim is that this is "decentralized" because no single party controls the funds.
The structural reality is different. A federation is:
- A custodian with extra steps. The federation members collectively hold your coins. You do not. If they coordinate, they can take everything. If a sufficient subset coordinates, they can take everything. The cryptography does not prevent this — it merely raises the coordination cost from "one person decides" to "M people decide."
- Trust-based, not trust-minimized. You are trusting that the M parties will not collude, will not be compromised, will not be subpoenaed, will not be coerced. None of these are cryptographic guarantees. They are reputational and legal guarantees, which is to say, the same guarantees offered by any bank.
- Opaque by design. Liquid's federation members are secret. Fedimint mints can have any membership their operators choose. In most cases you cannot verify who holds your funds. You cannot verify whether the "15 independent parties" are really 15 people, or one person with 15 keys, or three people each holding 5 keys, or a single corporation that compromised the original 15.
This is not decentralization. It is distributed custodianship, which is custodianship with extra failure modes.
The standard defenses, and why each fails
"But the federation can't run away with the funds without coordinating." True, and irrelevant. A bank cannot embezzle funds without internal coordination either. The coordination cost is not the security guarantee — the incentive and accountability are. With a federation, both are unclear.
"But Liquid has worked for years without incident." Past performance is not a security model. The same was said of every custodian that eventually failed, from Mt. Gox to FTX. The relevant question is not whether failure has occurred, but whether the system is structurally able to prevent it. Federations are not.
"But you can audit the funds on-chain." You can audit the balance of the multisig wallet. You cannot audit whether the keys are still held by the original parties, whether any of them have been compromised, or whether the off-chain accounting matches the on-chain reserves. This is the same audit problem that doomed every fractional reserve exchange in history.
"But the members have reputations to protect." Whose reputation? Liquid's federation members are not publicly disclosed. Fedimint guardians can be anyone the mint operator chose. Reputation only deters when the actor is identifiable, and most federations are designed precisely to avoid identification — supposedly to protect members from legal pressure, but the effect is to also protect them from accountability.
The crypto-economic test
Sztorc proposed a clean test for any L2 design: imagine the system at scale, hosting hundreds of billions of dollars in value, and ask what prevents the operators from stealing it.
For Bitcoin's L1, the answer is cryptographic: the discrete logarithm problem, SHA-256 preimage resistance, and the economic cost of acquiring 51% of global hashrate. None of these depend on the goodwill of named individuals.
For a drivechain sidechain, the answer is economic: miners are incentivized to honor withdrawals because they collect fees from sidechain activity, and the multi-month withdrawal delay gives the network time to respond to malicious behavior. The security is enforced by mining incentives, not by reputation.
For a federation, the answer is: nothing prevents theft except the federation's reluctance to steal. There is no cryptographic constraint, no economic penalty, no enforcement mechanism. If the members decide to take the funds and disappear, they take the funds and disappear. The protocol contains no method, no timing, no withdrawal-constraint, no alert system — just trust.
This is the model Bitcoiners used to mock Ethereum projects for using. "Dev keys" was a slur in 2017. Now it is marketing copy.
The brand problem
Layer this on top of Sztorc's observation about user trust models. Real users — the kind who will adopt Bitcoin in its eventual mass-market phase — do not trust protocols. They trust brands. They use Electrum because Electrum is a known brand. They use Trezor because Trezor is a known brand. The brand is a hostage that users can collectively execute by withdrawing their business if betrayed.
Federations break this model in the worst possible way. The user is told to trust "the federation" — an anonymous group with no public face, no consumer-facing brand, no recourse mechanism. When something goes wrong, who do you sue? Who do you boycott? Who loses their job?
Fedi's marketing is illustrative. The user is supposed to install the app and then "choose your federation" from a list. This is laughable as a consumer experience. Imagine a banking app that asked you to choose your bank from a list of anonymous LLCs. No one would use it. People will only use Fedi-style products if the federation choice is hidden from them — at which point the "federation" is just a backend implementation detail of what is, functionally, a custodian.
Why the federated model persists
If the federated model is structurally indefensible, why does it dominate the L2 conversation?
Because it is easy. Building a federation is a weekend project. You set up a multisig wallet, gather some respectable-sounding entities to hold keys, and ship. Building a real two-way peg is hard — it requires consensus changes, careful incentive design, years of review.
The federated model is what you build when you want to claim you have a sidechain without doing the work of actually having one. Blockstream's original sidechains paper, back in 2014, was clear about this: the federated model was explicitly described as a temporary alternative to the "true" two-way peg, suitable only for exchange-to-exchange transfers. Matt Corallo later put it bluntly: "if anyone tells you to use this for anything else, treat them like they're trying to steal your Bitcoin."
That warning has been comprehensively ignored. Liquid is now marketed as a general-purpose sidechain. Fedimint is marketed as a self-custody solution. The exception became the rule, because building the rule was too hard, and explaining the difference to users was too inconvenient.
What Ecash does instead
Ecash activates BIP-300 from genesis, which means the federated model becomes obsolete on day one — not by ban, but by competition.
A drivechain sidechain provides what federations claim to provide: a place to hold BTC outside the main chain, with arbitrary rules, without trusting a named operator. The difference is that the trust assumption is cryptographic and economic rather than reputational.
The cypherpunk ideal — trusting code, not people — is achievable. It just requires building the code, instead of assembling a committee.
Once drivechains exist, the marginal user has no reason to choose a federation. Why trust 15 anonymous parties when you can trust the mining incentive structure that already secures your L1 holdings? The federated model survives today only because the alternative is artificially unavailable. Make the alternative available and the federated model collapses into the niche it always deserved: short-term exchange-to-exchange transfers, where the trust window is brief and the parties are identifiable.
The honest version
I do not think most federation builders are dishonest people. I think they are people who wanted to ship something, looked at the political impossibility of activating drivechains on Bitcoin, and chose the second-best option that was actually deployable. That is a defensible decision as long as you call it what it is.
The dishonesty enters when federations are marketed as decentralized, when their members are hidden behind anonymity, and when the comparison to a real two-way peg is suppressed because the comparison is unflattering. The Liquid pitch deck does not say "we are a custodian with 15 cofounders." It says "decentralized sidechain." Those are different things.
Ecash's contribution is to make the honest version possible. A builder who wants to ship a sidechain on Ecash does not need to assemble a federation. They can use a real two-way peg. The dishonest marketing becomes unnecessary because the honest product is finally available.
That is what decentralization actually looks like. Not committees pretending to be protocols. Just protocols.