Symphony

The world remembers.

Who is looting the Mango DAO? Who is the mystery FTX buyer Max refuses to talk about? Why don’t Max and his cronies care?🧵 tldr: it was an inside job

@0xShitTrader You would find an equilibrium. Miners will only post proofs if it's profitable. If networking was fixed (the real problem right now), and scheduler was fixed, miners would naturally reach this equilibrium without an artificial controller.

Dynamic base fees would dramatically decrease the incentive to spam on Solana. Per-account dynamic write lock fees would quickly zero the incentive to mine $ORE, while returning that value to the network.

@MonetSupply Lack of jito mempool is not causing the UX issues. QUIC is.

bring back the jito mempool 🙏 ux hasnt been this bad since 2021 lol

@redacted_noah Terrible idea. mms and oracles would suffer. costs that would be passed on to retail no doubt.

Idea -- temporarily raise the Solana base fee to 0.0005 SOL (~$0.10). This will kill all of the spam, making the chain usable again until the scheduler is fixed.

@Justin_Bons @mert @cybercapital If prio fees are burned, users will just tip out of protocol instead

@mert That is a massive shame That will significantly downgrade SOL's score in @cybercapital evaluation models I am sure it will for others as well, if they are serious about economic design This makes SOL significantly less scarce into the future Will write a thread on this soon

@toly @mert The key point is to caution us from rushing out highly questionable features. We need more testing, research, and more time fto evaluate counters. We've already identified several issues in quic and banking stage that should be prioritized first.

@0xSymphony @mert > I don't agree with the "test in prod" mentality, that is how we end up with more outages. what is the liveness issue with 110? don't make me throw the CIA handbook at you

I can see what’s happening here You got things working on testnet and it’s great

@toly @mert SIMD 110 won't reduce tx load that much, if at all. These bots pricing out retail globally are bidding much higher for short lived opportunities that 110 cannot model or react to in time. The cause of latency and jitter in quic is not data packets.

@0xSymphony @mert less load, lower latency in the pipeline leading to the scheduler, faster prioritization. if it takes 100ms to get from the UDP kernel buffer to the scheduler, that tx is 100ms behind anything that is lower priority.

@toly @mert I agree with this, which is also scary because under 110 validators are also incentivized to jack up fees and form cartels.

@0xSymphony @mert Also, burned fee is basically not lost as validator income entirely. % staked * burned fee is effectively given to the stakers which is same as evenly distributed validator income. we are talking about 5% of value * 30% of unstaked tokens as actual lost validator income

@toly @mert Why would priority fee effectiveness be higher? All of the load is for short-lived opportunities where the 110 fees never have opportunity to rise in the first place. I don't agree with the "test in prod" mentality, that is how we end up with more outages.

@0xSymphony @mert if the average is 10x the median, maximum total value in the bottom half is 5%. So 5% gets burned for 50% reduction in load, and probably 10x in priority fee effectiveness. Only way to know is to test this on real traffic.

@SavvyShredder @mert @toly quinn needs to be replaced. Or even better quic needs to be eliminated. this needs to be #1 priority. ripatel's STL is a great start.

@0xSymphony @mert @toly What are you advocating as a better solution?

@toly @mert The extra fee gets burned. Either way, bots are only able to pay X value to the income is the same. Load on the quic server remains the same as well. Which is the real issue with validators having a fuller view of all txs.

@0xSymphony @mert or they gain more income because there is less load and prioritization is more effective. vast majority of the revenue is in the top half of txs. the delta between the average and the median is 10x.

@mert @0xren_cf 110 won't change their chance to be included. The connections are getting pruned at the quic level before they even reach banking stage

@0xSymphony @0xren_cf Yeah because people paying 1$ on jupiter today to have a totally stochastic chance of being included in a block today is so great

haven't done one of these in a while AMA about Solana and I'll do my best to answer or redirect

@mert @toly 110 will objectively make the chain worse. It doesn't solve spam. Several MEV opportunities arise and fall every single slot. 110 can only adjust base fees 1% every single slot. When demand falls, blockspace is left overpriced and valdiators lose income.

@mert @toly So you can't tell me how 110 solves the congestion problem?

@0xren_cf @mert Terrible. It will massively harm the chain

@mert what are your thoughts on SIMD-110?

@mert @toly Ok. I'll entertain it. What does 110 solve and how?

@0xSymphony @toly That is **a** problem, not the entire picture

@mert @toly Yes it is. The problem is the quic server. 110 doesn't fix it.

@0xSymphony @toly That's just not true

@mert @MF_JJJJ @toly Correct, the problem is the QUIC server, which supercedes the banking stage layer. Spam is still incentivized under 110.

@MF_JJJJ @toly are you saying 110 will have 0 effect on the chain experience right now?

@mert @toly SIMD 110 won't fix any of this

@toly I need you to enact an emergency executive order under the benevolent dictator act to enable 1.18 scheduler and simd 110 like literally tonight my family is dying