Gavin Wood - Mastering Ethereum

.

Fortunately, there were several safeguards built into The DAO: notably, all withdrawal requests were subject to a 28-day delay. This gave the community a little while to discuss what to do about the exploit, because from roughly June 17July 20 the DAO attacker would be unable to convert their DAO tokens into ether.

Several developers focused on finding a viable solution, and multiple avenues were explored in this short space of time. Among them were a DAO soft fork, announced on June 24, to delay DAO withdrawals until consensus was reached, and a DAO hard fork, announced on July 15, to reverse the effects of the DAO attack with an exceptional state change.

On June 28, developers discovered a DoS exploit in the DAO soft fork and concluded that the DAO hard fork would be the only viable option to fully resolve the situation. The DAO hard fork would transfer all ether that had been invested in The DAO into a new refund smart contract, allowing the original owners of the ether to claim full refunds. This provided a solution for returning the hacked funds, but also meant interfering with the balances of specific addresses on the network, however isolated they were. There would also be some leftover ether in portions of The DAO known as childDAOs. A group of trustees would manually authorize the leftover ether, worth ~$67 million at the time.

With time running out, multiple Ethereum development teams created clients that allowed a user to decide whether they wanted to enable this fork. However, the client creators wanted to decide whether to make this choice opt-in (dont fork by default) or opt-out (fork by default). On July 15, a vote was opened on carbonvote.com. The next day, at block height 1,894,000, it was closed. Of the 5.5% of the total ether supply that voted, ~80% of the votes (~4.5% of the total ether supply) voted for opt-out. One-quarter of the opt-out vote came from a single address.

Ultimately the decision became opt-out, so those who opposed the DAO hard fork would need to explicitly state their opposition by changing a configuration option in the software they were running.

On July 20, at block height 1,920,000, Ethereum implemented the DAO hard fork and thus two Ethereum networks were created: one including the state change, and the other ignoring it.

When the DAO hard-forked Ethereum (present-day Ethereum) gained a majority of the mining power, many assumed that consensus was achieved and the minority chain would fade away, as in previous forks. Despite this, a sizable portion of the Ethereum community (roughly 10% by value and mining power) started supporting the non-forked chain, which came to be known as Ethereum Classic.

Within days of the fork, several exchanges began to list both Ethereum (ETH) and Ethereum Classic (ETC). Due to the nature of hard forks, all Ethereum users holding ether at the time of the split then held funds on both of the chains, and a market value for ETC was soon established with Poloniex listing ETC on July 24.