[Bug]: CM charging bot fee when there's clock drift

alfongj commented 2 years ago

Which package is this bug report for?

candy-machine

Which Type of Package is this bug report for?

Rust Contract

Issue description

Steps to repro:

Set Candy Machine goLiveDate to X
Tell users "you can mint at time X! :DD"
Solana drift happens and Solana clock is 1hr behind real world clock
Users go to mint at time X and all they get is a lousy empty token and -0.01 SOL in bot fees

Impact: Seems decent, by looking at the amount of bot tax paid the days since the drift: https://dune.com/queries/860510

Possible solutions:

One option may be to use an oracle so that goLiveDate maps to real world time
Another option that sucks for users (because they won't know when a mint really opens) but doesn't depend on oracles is to update CM frontends to compare goLiveDate vs Solana time and not real world time
Other temporary band-aid solution would be to remove the bot tax from folks trying to mint before the goLiveDate. It's not clear to me that the bot tax on minting ahead of time is any deterrent to botters if there's already a bot tax on minting when the CM is sold-out

Relevant log output

No response

Priority this issue should have

Medium (should be fixed soon)

alfongj commented 2 years ago

Example of a recent Candy Machine:

In red: a spike of users who attempted to mint around 15 min of the goLiveDate, presumably because IRL / in the front end it was already time to mint

In green: requests when the mint actually opened

(I am not sure what explains the green requests pre goLiveDate, but they may be whitelisted mints)

https://dune.com/queries/860542

blockiosaurus commented 2 years ago

Hey @alfongj we definitely are tracking the occurrences on this issue, but due to the nature being derived from Solana's on-chain time our official stance is that all front-ends should use Solana's time as their source instead of off-chain time. This prevents the complexity of using an oracle and the backwards compatibility issues that that would cause. We've updated our front-end examples to reflect this stance, as well.

Solana's latest runtime release has also resolved the clock-drift issue, but feel free to re-open this ticket for further discussion should it occur again.

alfongj commented 2 years ago

Ok thanks. That was one of the suggested solutions on the OP.

Having said that this is still a bad UX if solaan drifts as users will need to know that launch dates will be delayed. But I get that solving it doesn't seem worth it if we think solana drift is not going to be that common

metaplex-foundation / metaplex-program-library