The eltoo proposal will make Lightning Network have improved functionality.
For you to understand how, first you will have to understand how LN channels work:
With the Lightning Network, users can transact via what are called payment channels. These channels allow parties to make an unlimited number of transactions between themselves for a specified period of time without broadcasting them to the bitcoin blockchain. When the parties are finished, they close the channel by transmitting the final updated status of the channel balance back to the main chain.
But what if a Lightning user tries to cheat their counterparty by broadcasting an outdated channel balance which would pay them more than they are due?
Lightning Network solves this by penalizing users who attempt to cheat by giving all of the funds they have open in the channel to the other party.
The problem is that the broadcasting of old balances is not always a cheating attempt. There are a number of scenarios in which users can accidentally broadcast an older balance; for example, because of a software bug or a backup gone wrong. In such scenarios, a complete loss of channel funds is quite a heavy punishment.
Eltoo updates a channel by building a chain of time-locked transactions, where each transaction spends funds from the previous one to reflect the latest channel balance.
This is basically makes the use of watchtower less important.
Eltoo continually updates payment channel balances by recording all the transactions, and then, at the time of settlement, allowing the intermediate states to be skipped so that only the last settlement transaction is confirmed to the blockchain. If one party tries to cheat by presenting an outdated transaction during settlement, the other side is given some time to transmit the latest transactions and thus ensure the accuracy of the channel balance.
This is a type of type of hierarchy in these types of transactions: any newer transaction can override any older transaction without requiring that all transactions in the entire chain be broadcast.
However, in order for eltoo to work on the Bitcoin network, a change to the Bitcoin protocol itself is necessary: the introduction of the SIGHASH_NOINPUT flag to signatures. This could be done via a soft fork, and the authors of the paper included a proposal to do so.
This works by constantly sending info to the opt code until the finalization sign is given closing the channel and recording it on chain. This is kinda works like smart contract where data is sent between until agreement is reached. In fact opt code is how Bch plans on getting its own version of smart contracts.