Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Mainnet Merge Announcement

blog.ethereum.org

Post reply on HN
Print view View original

451–460 of 609 posts

Re: Mainnet Merge Announcement

#451

Post by stavros »

squeaky-clean wrote: post #450
stavros wrote: post #444

Earlier quoted context omitted.

I am, yes. I don't think it's the memory because it segfaults immediately, even if I try 64x64... It's very odd, I'm running Ubuntu 22.04, I'm not sure if the CUDA version is somehow wrong.

Oh damn that's weird. Any chance the segfault is on the CPU side? When I try to generate too large of an image I get an allocation error rather than a segfault. Also crazy idea but could it be related to open source / proprietary driver differences? I haven't done nvidia on Linux in a long time so I don't know if that's in a more reasonable place yet.

Hmm, thanks for the info, it might be a CPU error! There's zero information in the error, it just says "error SIGSEGV" or something similar. I don't think it's an open source driver issue, I'm using nVidia's proprietary one...

It's so odd, it just dies outright. Maybe it is a CPU error, as you say, I think I'll investigate there.

Re: Mainnet Merge Announcement

#452

Post by rglullis »

spywaregorilla wrote: post #431
rglullis wrote: post #375

Earlier quoted context omitted.

You won't be prevent slashing with 1/3, you will be merely delaying the block production each time is is proposed. Every time your 1/3 of the nodes attempt maliciously delays of production of a valid block, you will be slashed, and then you won't have 1/3 of the required funds at stake anymore.

But... you can't be slashed unless a block gets processed that supports your slashing, correct?

Yes, but what is your point? You will be slashed no matter how hard you try to avoid it, unless you have 2/3 of the validators. The closer you are to 2/3, the faster you will be slashed. The system is self-estabilizing.

Re: Mainnet Merge Announcement

#453

Post by pa7x1 »

spywaregorilla wrote: post #431
rglullis wrote: post #375

Earlier quoted context omitted.

You won't be prevent slashing with 1/3, you will be merely delaying the block production each time is is proposed. Every time your 1/3 of the nodes attempt maliciously delays of production of a valid block, you will be slashed, and then you won't have 1/3 of the required funds at stake anymore.

But... you can't be slashed unless a block gets processed that supports your slashing, correct?

There is an inactivity leak that gets triggered if you prevent with over 1/3 of your votes from the network reaching consensus. Inactive validators start to leak ETH from their stake until they become less than 1/3 of the total stake and cannot prevent consensus again.

https://eth2book.info/altair/part2/incentives/inactivity

Re: Mainnet Merge Announcement

#454

Post by bowsamic »

camjw wrote: post #280
bowsamic wrote: post #260

Earlier quoted context omitted.

It’s cheaper to run 1 more GPU when you are already running 100 It’s the same price to add 1 eth when you own 100 eth as when you own 1 eth

Kind of, but doesn't that assume a linear utility function for holding eth which I think seems unrealistic? Plus, there are costs associated with holding eth (cold storage etc) which seem fixed, so its cheaper to securely stake 1 additional eth when you're already staking 100eth than when you're not running any. Plus fixed costs of actually running a staking business (by analogy with a mining business).

Yes, the utility function is linear for PoS, and wildly non linear for PoW

Re: Mainnet Merge Announcement

#455

Post by bowsamic »

pawelduda wrote: post #90
Anon1096 wrote: post #69

Earlier quoted context omitted.

32 eth is such a small investment that it isn't that meaningfully different than 1 eth.

Yeah, taking $4k (price it reached in the recent bull market). $4k vs $128k Literally the same numbers.

For the scales usually considered in the crypto scene these would be considered incidental

Re: Mainnet Merge Announcement

#456

Post by plq »

rauljordan2020 wrote: post #426
drog wrote: post #200

Earlier quoted context omitted.

There is no designated class of slashers. Slashing is part of the protocol and to slash somebody you have to prove it to the protocol that they broke rules of slashing. One of the rules is that you can't create two different blocks in the same slot (block number) i.e. you can't deliberately fork. There is a designated place in the block of the beacon chain where you can put signatures of the different blocks in the s…

I am one of the maintainers of the Go implementation of Ethereum proof-of-stake, called Prysm (github.com/prysmaticlabs/prysm) and also implemented a "slasher" in Go that can be used to slash malicious validators. Anyone can run a slasher and you don't need to have 32 ETH to do so. As long as your slasher software can prove that a validator committed a slashable offense, you can submit this proof to any full node to…

By "A few", how much do you mean? Or is that a dynamic number? How does it get defined? Who can change it later, if it's mutable?

Re: Mainnet Merge Announcement

#457

Post by mattnewton »

zionic wrote: post #198
erwinh wrote: post #151

What we see in the Tezos chain (liquid/delegated proof-of-stake) is that big custodial wallets for the exchanges have grown to be the largest block bakers: https://thestackreport.xyz/articles/top-tezos-block-producer... With ethereum the staking mechanism is a bit more complex, my understanding is you lock your stake for quite a while so maybe its too risky for the exchanges, but wouldn't be surprised that exchanges…

ETH PoS includes offline penalties proportional to the % of the network that goes down. If your home RasPi validator drops you don't lose much. If 25% of users stake on coinbase and _they_ go down the penalties are much higher. So there's an incentive to validate on your own hardware/connection.

If that were true why wouldn’t coinbase just stake through a few hundred different vms on servers instead of one block of nodes? Or is there some mechanism that ties it to a legal person / corporation?

Re: Mainnet Merge Announcement

#458

Post by michaelsbradley »

spicybright wrote: post #338
spywaregorilla wrote: post #240

Earlier quoted context omitted.

But all evidence needs to be voted on. Even it's just a divine truth bool, the network still needs to vote on a consensus of whether or not it says TRUE. edit: for an example. you get hacked and lose $100M. You say fuck it, submit a slasher report that says it wasn't a valid transaction. you offer enough money to get x% of validators to agree with you. As far as I can tell, you cannot get slashed for your vote on a s…

It's a shame you're being downvoted as your post has substance. This is a genuine issue worth discussing. Validators are run by people who fully control if a transaction is valid or not. You either need a ton of GPUs (currently) or a ton of money to have tangible influence over translations. This isn't even a hypothetical, there's a few massive wallets that are hoarding a ton of eth that will be given more power from…

> there's a few massive wallets that are hoarding a ton of eth that will be given more power from the POS transition

It's not as if they're automatically given more power. Such hoards of ETH would need to be deposited in units of 32 ETH to activate unique validators. You could have a large number of validators that are actually a single computer program participating in the Beacon chain, but that doesn't make too much difference.

At the bottom layer, we're talking about the application of an open p2p protocol on the Internet. People can and will analyze the traffic for that protocol, and it will be possible to identify rough shapes and sizes of players in the network. And such analyses won't be long in coming to public blog posts, etc. following the Merge.

If decentralization in the network seems at risk, there will be efforts to remedy the situation. Whether they can succeed is another question.

Re: Mainnet Merge Announcement

#459

Post by tingletech »

factorialboy wrote: post #440
aliljet wrote: post #3

I've been eagerly waiting for this to happen for a while. Here's picking up GPUs at fire sales for months and maybe years to come.

Isn't the whole point that outrageous processing power is no longer a prerequisite.

right, so presumably there will be a lot of idle GPUs that miners will sell / less demand for new GPUs?

Re: Mainnet Merge Announcement

#460

Post by overtonwhy »

zaptrem wrote: post #33
spywaregorilla wrote: post #19

Earlier quoted context omitted.

This is not the question I am asking. You can have many people putting up their 32 ETH. Sure that's "decentralized". But what if most of the stakers collude to double spend tokens? My understanding is that there is an additional layer of validators that have the power to force them to give up their stake as a penalty. Hence "proof of stake". And that there are very few validators. And it's really just the ethereum fo…

You’re mistaken, validators and stakers are one and the same. A 51% attack becomes unbelievably more expensive in a proof of stake system, especially because at the end of it all of your money is burned (whereas with POW you get to keep the GPUs unless they change the algo).

What about a DDoS attack?
Post reply on HN