Given that when Fabien Vogelsteller launched LUKSO he used an ERC777 and not an ERC20 which he was a co-author of, it seems weird that the ERC777 should be removed from the OpenZeppelin libraries.
It may not be as popular as the ERC20 but I cannot even see any explanation as to its removal anywhere. The only arguments that I have heard is that it is possible to leave vulnerabilities if you use it wrongly - but that same argument can be applied to ERC1155 or 721's safetransfer.
The ERC777 contract was deprecated in 4.9 but not removed until 5.0
We did it this way for backwards compatibility reasons.
Regarding the reasons behind the removal, the ERC777 standard was created with the motivation of extending an ERC20 with hooks. This was allegedly in order to provide users more control over their ownership (eg. bypassing
approve/transferFrom for a single hooked
However, the hooks system in practice (among other features) became difficult to reason about and very error prone.
The potential vulnerabilities (although similar to ERC1155's and ERC271's cases) not only have to do with reentrancy when calling the acceptance check on the recipient (
tokensReceived) but with reentrancy when calling the
tokensToSend hook since it breaks the check-effects-interactions pattern.
It was solved with ERC1820 registry but at that point there was some community consensus that ERC777 was becoming over-engineered, a source of frustration, and better alternatives came out. You can see more of the discussion here.
Currently, the new
_update mechanism in v5's ERC20 is more than enough for extending capabilities in a secure way
Very annoying for projects that implement ERC777 for completely valid reasons and have taken care to do so correctly.
Can you elaborate on these valid reasons to use ERC777?
Perhaps a project using an ERC777 may find an alternative in other standards/solutions but I'd be interested in hearing if the removal left some needs unmet.
The issue is not with the project that implement tokens using ERC-777, its with the other projects/contracts that will interract with these tokens in a potentially dangerous way ... and all the user of account abstraction that have to deal with ERC-1820 registration.
So we need to protect people who cannot code properly?
Perhaps we should remove ERC1155 for the same reason and get rid of safeTransfer in ERC721?
When people send ETH they face exactly the same issues - that the contract that they are sending to can
- refuse the ETH
- commit a re-entrancy attack
This is why we have re-entrancy guard and teach people to update the state before sending. Nobody would argue that the receive function is a bad thing. ERC777 offers something similar to the ETH receive function.
Failing that, implement ERC677 so that we have an audited specific transferAndCall mechanism that can be used without the call being trigered on standard ERC20 operations.
This is even worse - we introduce a new standard that is problematic because we have an existing standard that makes it hard to implement and so junk the old standard and ignore its existence regardless of any old tokens that may be out there?