Chronos: A Quirky Application Proposal for Plasma

[PaulRBerg]

Thanks so much for thinking this through! It’s a really helpful discussion.

@danrobinson

Since it also probably requires an on-chain transaction every time a stream is opened (unless Alice happens to have the right denomination to open it), why not just do the stream logic in a smart contract on the mainchain?

I guess you’re referring to something like this:

https://github.com/ChronosProtocol/poc/blob/master/contracts/Stream.sol

The rationale for choosing the infrastructure as proposed in the OP over an on-chain smart contract is:

1. Fast finality via a single operator/ PoS set of validators, getting rid of reorgs and uncles

2. Better time measurement, due to shorter interval blocks

3. Cheaper state updates

@danrobinson

The mainchain contract knows the early close time (the time when the attempted withdrawal of the coin happens on the main chain). So it can just do the math.

Hmm you’re completely right! I was unconsciously imagining the Plasma contract as a black box, probably because of this excerpt:

@danrobinson

I’m not sure it would work to use a ZK-snark for the non-existence proofs, unless the parent chain is willing to accept ZK-snarks for the exit transactions. Otherwise, you don’t have enough information to respond to a challenger, right?

zk-SNARKs are a different beast though. For streams, you can just do the difference between the closing and starting times and multiply by the rate - and indeed, the closing block could be the withdrawal attempt time. The chains have different block numberings, but you could potentially agree on a rough estimation of time/value of money if you keep a block height scaling coefficient in the contract (and update that accordingly over time, similar to how retargeting works in PoW).

The only problematic edge case I see is when the rate is calculated at smaller intervals than the block time on the main chain. That is, if the agreed off-chain rate is x units per y seconds and we consider the main chain to have z seconds blocks, where z>y by a non-trivial margin, the potential loss could be as high as (z-y)*x (“loss” as in Bob would get that money instead of Alice, because she tried to exit exactly y seconds after a block was generated on the main chain). I admit though that this potential loss is negligible in most cases, especially when building an MVP.