4 ms·
It seems like a lot of these problems are due specifically to using proof of work. Do any of these problems go away with other consensus algorithms? I would at
by scyclow 9y ago
It seems like a lot of these problems are due specifically to using proof of work. Do any of these problems go away with other consensus algorithms? I would at least expect that miners wouldn't have as much incentive to arbitrarily cause errors if they were using something like proof of stake.
Also, how easy is it to detect that miners are doing these things? If it's not terribly difficult, then maybe a federated system would iron out some of the wrinkles?
- mhluongo 9y agoThey do! But usually they get shifted elsewhere. While he was working on this post James and I discussed how PoS might yield stronger incentives to DoS rivals and get an advantage, since some of these other optimizations (attacks? Unclear) will be unavailable. Edit: Regarding detection, of course- but I'm not sure detection lends itself to a general solution outside name-and-shame. Most of these are things developers should code defensively against. Disclosure: lead at Keep, edited this piece
- prestwich 9y agoHello, author here. I haven't put a lot of thought into this yet, but my first guess is that this could become more of a problem with staking. Stakers still need to optimize, but all the easy ways are gone. Can't build better ASICs for staking. Detection is hard. Like with selfish mining, it looks just like normal network operation in most cases. Most obvious mitigations don't work well either. E.g. we could try to establish canonical tx ordering in the tx pool. But then we would have to come to consensus about what the txpool is. Which is a bit of a chicken/egg problem.