2 Existing Work
Decentralized exchanges implemented with Ethereum smart contracts have failed to generate significant
volume due to inefficiencies in their design that impose high friction costs on market makers. In particu-
lar, these implementations place their order books
1
on the blockchain [5–8], requiring market makers to
spend gas each time they post, modify or cancel an order. While the cost of a single transaction is small,
frequently modifying orders in response to evolving market conditions is prohibitively expensive. In addi-
tion to imposing high costs on market makers, maintaining an on-chain order book results in transactions
that consume network bandwidth and bloat the blockchain without necessarily resulting in value transfer.
Automated market maker (AMM) smart contracts are proposed [9,10] as an alternative to the on-chain
order book. The AMM smart contract replaces the order book with a price-adjustment model in which
an asset’s spot price deterministically responds to market forces and market participants on either side of
the market trade with the AMM rather than with each other. Benefits of the AMM include availability
(it is always available to act as a counterparty, though the spot price it offers may be worse than what
one could get from a more traditional exchange) and ease-of-integration with external smart contracts
that need to execute market orders. The deterministic nature of price-adjustment models make them
insensitive to market liquidity, meaning that trades cause prices to move the same amount in both thick
and thin markets [11]. In other words, AMMs impose artificial constraints on the supply curve. If the
price-adjustment model is too sensitive, even small trades will produce large fluctuations in the spot
price. If the price-adjustment model is not sensitive enough, the AMMs bankroll will quickly be depleted
by arbitrageurs.
State channels are proposed as a means of scaling the Ethereum blockchain and reducing costs for a
variety of applications - including exchange [12] - by moving transactions off of the blockchain [13, 14].
Participants in a state channel pass cryptographically signed messages back and forth, accumulating
intermediate state changes without publishing them to the canonical chain until the channel is closed.
State channels are ideal for “bar tab” applications where numerous intermediate state changes may be
accumulated off-chain before being settled by a single on-chain transaction (i.e. day trading, poker,
turn-based games). If one of the channel participants leaves the channel or attempts to cheat, there is a
challenge period during which the other participant may publish the most recent message they received
from the offender. It follows that channel participants must always be online to challenge a dishon-
est counterparty and the participants are therefore vulnerable to DDOS attacks. While state channels
drastically reduce the number of on-chain transactions for specific use cases, the numerous on-chain
transactions and security deposit required to open and safely close a state channel make them inefficient
for one-time transactions.
A hybrid implementation, which we refer to as “off-chain order relay with on-chain settlement,” com-
bines the efficiency of state channels with the near instant settlement of on-chain order books. In this
approach, cryptographically signed orders are broadcast off of the blockchain; an interested counterparty
may inject one or more of these orders into a smart contract to execute trades trustlessly, directly on the
blockchain [15, 16]. Friction costs are minimized for market makers because they can signal intent off-
chain and transactions only occur when value is being transferred. We extend this approach by allowing
anyone to act as the exchange and by making the protocol application-agnostic.
1
An order book is used to publicly record the interest of buyers and sellers in a particular financial instrument. Each entry
includes a reference to the interested party, the number of shares and the price that the buyer or seller are bidding/asking
for the particular security.
4
评论0