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

The Merge

ethereum.org

Post reply on HN
Print view View original

241–250 of 414 posts

Re: The Merge

#241

Post by landemva »

tzs wrote: post #153
bhaak wrote: post #112

Earlier quoted context omitted.

All the nodes that have been online long enough will know that the fake chain is fake. Because it doesn't match the state that they have been observing all along. So as long as you have no extended downtime, your nodes know what the right chain has to look like. Ethereum dynamically adds checkpoints so that block reorgs can't reach too far into the past but that is mostly a convenience function as it alone couldn't s…

Could POS and POW be combined? Say you have a POS chain that functions as your main cryptocurrency ledger, and a POW chain that runs at a much slower rate that serves as a ledger to periodically record history of the POS chain?

Sort of, if btc checkpoints eth. Hash the entire eth PoS state tree every hour or so and include that hash with block height in a bitcoin transaction.

Re: The Merge

#242

Post by thesausageking »

landemva wrote: post #238
thesausageking wrote: post #199

Earlier quoted context omitted.

PoS is less proven than PoW and has some edges cases that give many concerns. To say this up front, I own ETH, have worked on dapps on it, and completely supprt the move to PoS. That said, the fact that PoW has been running the longest and Bitcoin uses it to secure the most value of any chain and it has had 100% uptime for the last 9.5 years, means people naturally trust it more. PoS also in theory incentivizes centr…

> 100% uptime for the last 9.5 years While correct, that conveniently omits the downtime in 2010 and 2013. Why be shady? Edit: three nines is considerably different than infinite nines. Around 2018 there was an inflation bug which was kept quiet until it was patched. It's human to have an occasional bug. Setting the bar too high may be a dis-service if/when there is an exploit.

What? How is that shady?

It had 99.98% uptime since launched and 100% since 2013. That’s one of the most reliable software projects ever created in history.

Re: The Merge

#243

Post by rakoo »

miohtama wrote: post #154
rakoo wrote: post #56

Earlier quoted context omitted.

Thanks for the explanation! I have a few questions: - how does the punishment work ? Do all nodes see that a malicious actor changed the chain and write the bad ideas in some ledger meaning "these nodes just lost X ETH" ? - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted - is there an advantage to run more than one validator, i…

You can find more details on slashing and penalties in the Validator FAQ: https://launchpad.ethereum.org/en/faq

Wow thanks a lot !

Re: The Merge

#244

Post by rakoo »

hjorthjort wrote: post #180
rakoo wrote: post #56

Earlier quoted context omitted.

Thanks for the explanation! I have a few questions: - how does the punishment work ? Do all nodes see that a malicious actor changed the chain and write the bad ideas in some ledger meaning "these nodes just lost X ETH" ? - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted - is there an advantage to run more than one validator, i…

An important and often misunderstood point: Having >2/3rds of validators (or >50% of power in PoW) let's you rewrite the chain back in history, censor txs, etc. You control what transactions go in blocks, and to a certain extent also recent history. HOWEVER, this does not give you the power to steal or rewrite state as you please. You cannot convince the rest of the network that you now own someone else's ETH just be…

Well this is the same as any blockchain: you can't forge a block with a transaction giving yourself money, but you can double spend the same amount to multiple other accounts. Unless something is different with ETH which I don't know enough

Re: The Merge

#245

Post by rakoo »

entrep wrote: post #88
rakoo wrote: post #56

Earlier quoted context omitted.

Thanks for the explanation! I have a few questions: - how does the punishment work ? Do all nodes see that a malicious actor changed the chain and write the bad ideas in some ledger meaning "these nodes just lost X ETH" ? - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted - is there an advantage to run more than one validator, i…

> - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted As I understand it, anyone who owns that amount of ETH would probably be interested in contributing to the trust of the Ethereum blockchain.

Call me a pessimist, but I have stopped assuming rationality of all actors acting to secure a long-term future, especially in the world of cryptocurrencies

Re: The Merge

#246

Post by cypress66 »

rakoo wrote: post #244
hjorthjort wrote: post #180

Earlier quoted context omitted.

An important and often misunderstood point: Having >2/3rds of validators (or >50% of power in PoW) let's you rewrite the chain back in history, censor txs, etc. You control what transactions go in blocks, and to a certain extent also recent history. HOWEVER, this does not give you the power to steal or rewrite state as you please. You cannot convince the rest of the network that you now own someone else's ETH just be…

Well this is the same as any blockchain: you can't forge a block with a transaction giving yourself money, but you can double spend the same amount to multiple other accounts. Unless something is different with ETH which I don't know enough

Double spends are actually much harder here, after slots are finalized (2 epochs, about 12 minutes) but yes it's the same as other blockchains.

Re: The Merge

#247

Post by ChadNauseam »

DennisP wrote: post #192
SilasX wrote: post #160

Earlier quoted context omitted.

What happens if most validators become non-responsive without unstaking? The network can’t create new blocks then?

The network would still create new blocks but wouldn't finalize them. Finalizing means they can't be undone without destroying a large percentage of the stake. Without finalization, you still reach a point where it's very unlikely that the block will be undone. The non-responsive stakers would gradually lose their stake. The penalty for going offline is normally small, but if most validators go offline at once it's c…

Finalized blocks can never be undone, it’s forbidden by the protocol. Once your client sees that a block is finalized, it will never revert it.

(This makes the network a bit subjective, so hopefully it’ll come up rarely)

Re: The Merge

#248

Post by ChadNauseam »

tootie wrote: post #216
randomran01234 wrote: post #211

Earlier quoted context omitted.

It takes time to learn but it isn’t obtuse. It’s extensively documented, with multiple open source clients being built in tandem. Besides, it is not the only PoS system running today. All fields of engineering are complex to those that have an not studied them.

The mechanics aren't extremely complicated but they also make no sense. Proof of work or proof of stake are both solving an abstract and meaningless problem. Crypto makes more sense if you study psychology than engineering.

Permissionless BFT consensus is definitely an engineering problem

Re: The Merge

#249

Post by rakoo »

WFHRenaissance wrote: post #117
rakoo wrote: post #56

Earlier quoted context omitted.

Thanks for the explanation! I have a few questions: - how does the punishment work ? Do all nodes see that a malicious actor changed the chain and write the bad ideas in some ledger meaning "these nodes just lost X ETH" ? - But if the malicious actor owns 2/3 of validators, what does he care what other nodes do ? They effectively control what block is accepted - is there an advantage to run more than one validator, i…

1. Punishment mechanisms here are kind of complicated, but in short, you're punished for mainly 2 reasons: being an offline validator, or for an attestation violating. Slashers are entities that enforce the two above rules. If a slasher determines that you're node is down, or that you're committing an attestation violation (i.e. that you're signing more than one attestation in a given epoch mainly). The slasher actua…

Thanks for the reply !

1. So if I get it correctly, there is an actor in charge of surveiling the network. There is no financial incentive to do that which means only a few actors will be running a slasher, and thus completely remove all decentralization. Once a fraud has been detected, the proof of fraud is to be spread by nodes with a financial incentive, meaning they might spread whatever the slasher says like gospel, giving even more power to slashers.

This is a very large step away from decentralization. The network depends not just on a few coders, but also on an institution checking whether nodes behave correctly or not. The whole point of blockchain was that everyone would assume everyone was out to cheat, and came up with a way to make it useful.

2. As I said in another reply, I don't expect the most rationality and long-term thinking from economic actors anymore, especially in the world of cryptocurrencies. I could very well see a single actor owning that many nodes but still being trusted because "they're the good guys" and people remain. You would also have to be sure they are in fact colliding.

3. So, as I feared, this is capitalism in its crudest form. At least PoW made it indirect by putting requirements on the hardware, but here it's just plain naked for all to see.

I know I sound very negative but I was really curious to see how PoS really changed the dynamics compared to PoW, and now I know. Thanks again for your explanations.

Re: The Merge

#250

Post by ChadNauseam »

giaour wrote: post #156
pkulak wrote: post #111

Earlier quoted context omitted.

Nothing. But nothing moves forward without consensus. So you'd need 51% of the network to agree with your made up chain, which would mean controlling about $100 billion in Eth. And then you've managed to destroy a network that you have $100 billion wrapped up in, so... good job? And if you try with less than 51% and the rest of the network slaps you down, you lose your stake (or some portion of it).

> So you'd need 51% of the network to agree with your made up chain, which would mean controlling about $100 billion in Eth. Does that sum assume that every coin will be staked? As I understand it, coins staked for validation are held in escrow and can’t be used for transactions until the validator node unstakes the coins. Surely only a small percentage of coins will be staked at any given time.

There is $200B ETH staked already
Post reply on HN