Decentralisation is key to Nu success, and I am glad to see new custodians supporting Nu. With more custodians joining the network in the next weeks, managing markets becomes increasingly more challenging. Especially in terms of collisions among bots on the same market.
We are working hard to make it possible to have multiple custodians operating NuBot on the same pair. Testing on our internal exchange showed good results.However, its not realistic to reproduce the reality of markets in a testing sandbox.
As one of the NuBot developers, I need to be able to test NuBot in the production scenario and quickly react to avoid collisions.
Custodian: Desrever
Submission Date: 21/11/2014
Requested Grant Amount: 600 NBT
NuBits Grant Address: BCustdn3aS3oU9FXKbjJ3mphcpru5dUxAh
Vote Amount: 600 NBT
This is really important, folks. I would strongly urge everyone to vote for this.
It can be very difficult for the developers to isolate issues with the bot without being able to see it first hand. Only custodian addresses are allowed to send liquidity info to the client, which is why desrever will need the grant. Also, it is unfair to ask him to test with his own funds.
I will use my own funds, but use the grant as an insurance to cover the fees of collisions :
This operation puts both my own funds and other custodians funds at risk of collision. However, the size of my walls will never be greater 1000 NBT. So each collision, if any, will cost approximately 2 NBT (0.2 % fee).
In case of a collision happens I will verify the causes, react, and revert the transaction where possible.
…
Upon receiving the 600 NBT of the grant, they will be parked for approx 20 days. When the unpark transaction happens the funds will be first used to refund custodians (including myself) in case they had to pay a fee on the collision. These fund is enough to cover ~ 300 collisions, 15 per day.
According to the proposal I should park the grant for 20 days, but the client won’t let me since there is no interest rate for that duration ( 0% ) .
Important update : the first hours of testing yesterday immediately revealed a design flaw in the multi-custodian system. It will probably require a new design and thorough testing before we can release another candidate. Therefore I will suspend any activity of testing and count only 1/20 of the contract filled.
See? This grant was indeed useful, the flaw is subtle to spot but gross.