EtherDelta

Summary

EtherDelta, launched in July 2016 by U.S. developer Zachary "Zack" Coburn, was the first Ethereum DEX to achieve meaningful adoption. It let anyone trade ETH against any ERC-20 token without giving custody to a company: funds sat in an audited-by-crowd, open-source smart contract, while a website run by Coburn hosted the order book. During the 2017 ICO boom it became the default secondary market for freshly issued tokens — users placed more than 3.6 million ERC-20 token orders in roughly 18 months, and at its late-2017 peak it handled daily volumes in the tens of millions of dollars. The project then unraveled in quick succession: a DNS-hijack phishing attack in December 2017, a sale to opaque new owners whose token ICO fell far short of its goal, and a community fork (ForkDelta) that kept the original contract alive after trust in the operators collapsed.

Design (Mechanism)

EtherDelta was a hybrid: on-chain custody and settlement, off-chain order discovery.

• Deposit contract with internal balances. Users deposited ETH or ERC-20 tokens into the EtherDelta contract (the long-lived version is EtherDelta_2 at 0x8d12...6819, ~300 lines of Solidity, verified on Etherscan and open-sourced on GitHub). The contract tracked per-user, per-token balances; trades were internal balance swaps, so a trade cost far less gas than token transfers.
• Signed off-chain orders. A maker constructed an order (tokenGet, amountGet, tokenGive, amountGive, expiry block, nonce) and signed its hash. Orders were broadcast via EtherDelta's centrally hosted order book server and website (an on-chain order() posting path also existed but was rarely used due to gas cost).
• Taker-executed settlement. A taker called trade() with the maker's order and signature. The contract verified the ECDSA signature, expiry, and remaining fill, then atomically adjusted internal balances. Partial fills were supported; makers could cancel on-chain.
• Fees. Takers paid 0.3% to a feeAccount controlled by the operator; makers paid nothing, so the taker-pays model gave makers free limit orders.
• Permissionless listing. Any token could be traded by pasting its contract address — no listing approval. This made EtherDelta the instant secondary market for every new ICO token, and also a playground for counterfeit tokens with copied tickers.

The trust model was deliberately lopsided: settlement and custody were trustless, but order relay, the UI, and the DNS record were a single centralized point of failure.

Outcome

EtherDelta rode the 2017 ICO wave to become the dominant DEX of its era, then failed on every non-contract layer:

• December 2017 DNS hijack. An attacker socially engineered the domain registrar, pointed etherdelta.com at a pixel-perfect clone, and harvested private keys. At least 308 ETH (~$266K at the time) plus tokens were stolen in the initial window, with total losses across the incident later estimated near $1.4 million. The smart contract itself was never breached — hardware-wallet and MetaMask users were unaffected.
• Sale to new owners. Coburn sold EtherDelta to foreign buyers in late 2017. The new owners launched an EDT token ICO that reportedly reached only ~3% of its hard cap. Community trust collapsed, and in early 2018 volunteers forked the open-source client as ForkDelta, pointing at the original contract.
• Obsolescence. By 2019, 0x relayers, IDEX, and above all Uniswap's AMM model had absorbed the user base. The contract still exists and holds stranded funds, but the platform is defunct; the surviving etherdelta.com/related properties are widely treated as unsafe.

Why it worked

• Non-custodial settlement was a genuine breakthrough after Mt. Gox-style failures: the operator could not steal deposits, and the contract survived every attack on the platform.
• Permissionless listing matched the moment. Every ICO token was tradable minutes after deployment, giving EtherDelta a monopoly on the long tail during the 2017 boom.
• The internal-balance design was gas-efficient and the taker-pays fee model gave makers free limit orders, seeding liquidity.

Why it failed or underperformed

• Centralized everything-else. DNS, the order book server, and the UI were single points of failure run by one person; the December 2017 hijack proved "non-custodial" does not mean "trustless end to end."
• Governance by sale. Handing the platform to anonymous buyers mid-crisis destroyed trust faster than any exploit; the community fork captured the remaining goodwill.
• UX and adversarial listings. Clunky deposit/withdraw flows, tokens lost to direct transfers, and counterfeit-ticker scams made it hostile to newcomers, leaving it defenseless once Uniswap offered one-click swaps.

Lessons

• Decentralizing custody but centralizing order flow, DNS, and the frontend just relocates the attack surface; users experience the weakest layer, not the strongest.
• Open-sourcing the stack is real resilience: ForkDelta could resurrect the service against the original contract the moment the operator lost legitimacy.
• Permissionless listing is a growth superpower and a fraud vector at once; without namespace protection (verified addresses, not tickers), scams tax every user.
• First movers with poor UX have no moat: the AMM's abstraction of the order book erased EtherDelta's incumbency within months.

Redesign (EDITORIAL — hypothesis, not fact)

This section is editorial hypothesis, not historical fact. A redesigned EtherDelta would keep the settlement contract but decentralize the relay layer: orders gossiped over a p2p network or posted to multiple independent relayers (as 0x later attempted), with the frontend deployed on IPFS/ENS to neutralize DNS hijacking. Fees would accrue to a contract governed by token holders or be burned, removing the single-operator dependency entirely. Listings would be keyed to verified contract addresses with on-chain provenance metadata rather than tickers. Finally, an explicit succession mechanism — a time-locked, community-ratifiable handover of domain, keys, and fee switch — would have prevented the trust collapse that the opaque 2017 sale caused. Even so, the deeper lesson stands: passive liquidity (AMMs) beat maker-taker order books for long-tail assets, so the redesign would likely still have lost the market it created.