MaxResnick1
0 Following
0 Followers
0 Liked
0 Shared
Posts
Fartcoin is digital liquified natural gas
Hot air always rises
Hot air always rises
On static transactions:
Transactions should be fully static at submission.
Why: Sequencers must be able to reason about transactions without executing them. And validators should be able to reason about block validity without executing the block.
All sequencing-relevant inputs including but not limited to:
1. Fee Payer
2. Resource usage and fees charged
3. Insructions
4. Read and write access list
Should be contained within the transcation.
In fact, even which accounts are loaded should be static and not manipulable in the runtime to avoid cache misses (you certainly should be able to not read something that you asked for but you should not be able to read something that you didn't ask for).
If the sequencer can’t simulate your transaction’s account locks and resource use without access to the runtime, its too dynamic.
Transactions should be fully static at submission.
Why: Sequencers must be able to reason about transactions without executing them. And validators should be able to reason about block validity without executing the block.
All sequencing-relevant inputs including but not limited to:
1. Fee Payer
2. Resource usage and fees charged
3. Insructions
4. Read and write access list
Should be contained within the transcation.
In fact, even which accounts are loaded should be static and not manipulable in the runtime to avoid cache misses (you certainly should be able to not read something that you asked for but you should not be able to read something that you didn't ask for).
If the sequencer can’t simulate your transaction’s account locks and resource use without access to the runtime, its too dynamic.
"Why is p90 wait time on ping thing still so long?"
I think most people (who work on fee markets) understand that wait time in a FIFO m/m/1 queue with excess capacity is long tailed. That is most of the time you arrive when the queue is empty but some small % of the time the queue is full and when its full the wait time is super long in expectation when this happens.
What people might not realize is that this is also the case for priority queues for any given priority level. Supose you are paying a priority fee that is higher than 30% of the other incoming traffic. Then your wait time in the priority queue is at least as long as under the FIFO queue with 70% load, and probably higher and much longer tailed because if you are sitting in the queue, any transactions with a higher priority fee that arrive go before you.
I think most people (who work on fee markets) understand that wait time in a FIFO m/m/1 queue with excess capacity is long tailed. That is most of the time you arrive when the queue is empty but some small % of the time the queue is full and when its full the wait time is super long in expectation when this happens.
What people might not realize is that this is also the case for priority queues for any given priority level. Supose you are paying a priority fee that is higher than 30% of the other incoming traffic. Then your wait time in the priority queue is at least as long as under the FIFO queue with 70% load, and probably higher and much longer tailed because if you are sitting in the queue, any transactions with a higher priority fee that arrive go before you.
Solana is not focused on CLOBs, FCFS, cancels before takes, or any other market structure. There are no Zealots here, only pragmatic people who want the tightest most liquid markets to exist on Solana. Whichever market structure leads to that is fine with us. But the reality today is that if an app wants to expirement with some new market structure Solana isn't the place to do that. Applications don't control their own sequencing on Solana, Validators do. That is not going to work long term.
The lion does not concern himself with the underlying market structure he is trying to build his defi app on.
Live footage of people who said blockchains couldn’t scale, and that nobody would use the SVM because the EVM moat was too big claiming that a general purpose blockchain cannot have application specific sequencing.
The idea that a general purpose blockchain cannot have application specific sequencing is now being pushed by the very same geniuses who told you that blockchains couldn’t scale.
Solana proved them wrong before and we are going to prove them wrong again.
Solana proved them wrong before and we are going to prove them wrong again.
Idk who needs to hear this but an oracle update is also a cancel when the AMM quotes around the oracle update.
Has anyone done user research on how liquidation impacts user retention on perps. Seems like there is a simple regression discontinuity aproach comparing users who almost got liquidated to users who did.
Either you sort the transactions by priority fee before they go into the block or you sort them during replay...
Y’all are in my hood this week so you best behave yourself
American DeFi
American DeFi
Async Execution takes Execution off the critical path but it doesn't eliminate the need for blocksize limits. In fact, blocksize limits are even more important to get right under Async than without it. You still need to know that the block will be replayed in time.
If you are a Solana validator attending Solana Accelerate, we are hosting an exclusive event for validators that you will not want to miss. DM me for invite.
Verifiable finance is important because when you throw away your entire life savings on a memecoin you need to be 100% sure that is what happened and you can verify it on your own hardware.
Login to explore more contents