Quote from: iamnotback on November 12, 2016, 12:24:55 PM

Seems I am banned from Reddit, because the following posts have been removed:

Quote from: https://www.reddit.com/r/Bitcoin/comments/5cerbi/researchers_suggest_new_method_to_scale_bitcoin/d9x35zi/

Afaics, Byzcoin can't possibly fix the most fundamental flaw in Satoshi's design when block rewards come predominantly from transaction fees, which is that the transaction fees either decline to mining costs or the throughput must be limited (e.g. by block size) so that transaction fees must rise to those the larger valued transactions are willing to pay.

One might argue that some transactions are willing to pay a higher transaction fee to be included sooner in a block, yet without any restriction on throughput (e.g. block size), this merely means one has to pay a transaction fee that makes the marginal miner profitable. But there is no such fee, because the marginal miners aren't just one level of cost. Thus gradually the most marginal miners go bankrupt, which proceeds until the lowest cost miners control 51% of the network and can raise fees to what ever level the market will bear.

The marginal miners rejoin the network if the cartel raises the level of transaction fees accepted but this is only stable if there remains a 51% cartel to raise fees. Due to the economics of Satoshi's design (e.g. minority mining on the wrong block during propagation delay, selfish mining, and even stubborn mining), those with more hashrate earn more profit than their proportional hashrate should, thus the 51% cartel over time trends towards 100% and so as their percentage of the systemic hashrate rises, the cartel can raise transaction fees to higher levels up to what the market can bear. So eventually we will be right back at Visa and Mastercard levels of centralized control and fees.

The "greedy mining attacks" are essentially a manifestation of the same underlying economic problem which is that given no restriction on throughput, then game theory incentives cause transaction fees to decline to the mining costs. Byzcoin may fix some of these attacks, but it can't fix the fundamental problem with Satoshi's proof-of-work, because it is insoluble. And no, Monero didn't fix this problem as TPTB_need_war explained to ArticMine earlier this year.

I (as @AnonyMint) had pointed out back in 2013 that transaction fees are the Achilles heel of Satoshi's design.

Ultimately what this all means is that mathematically and microeconomically for Satoshi's design to survive, a cartel must control a monopoly on mining (e.g. 51% of the hashrate) so that it can dictate a level of fees which is profitable. This is potentially why the Chinese mining "cartel" has been afaik resisting block size increases, because at least this is more obfuscated and more immediate than a battle of attrition or 51% attack to rid the blockchain of miners not in the cartel for the purpose of increasing profit from transaction fees. And there may be other vested interests in preference for Lightning Networks.

There can't ever exist any solution for Satoshi's proof-of-work design that will prevent a devolution into a mining cartel.

The only way to improve on this might be to shift to a design which doesn't use proof-of-work with blocks.

Byzcoin is also flawed because the splitting of the fees between all of those in consensus block can be subverted by offering lower fees to transactions which pay a well-entrenched miner out-of-band, thus the blockchain doesn't know that those fees were supposed to be split up. Paul Sztorc is correct here, and Emin Gün Sirer is incorrect.

Emin Gün Sirer is trying to argue that other mines won't mine on the block (or will reject its transactions) which doesn't fully share fees, but there is no way to objectively determine this since the offending transactions can still include a small fee (that is shared) in addition to the (discounted!) fee paid in the side band. Also Emin Gün Sirer is referring to Bitcoin-NG there, but the discussion is really about Byzcoin. In Byzcoin a miner with for example 25% of the hashrate is going to be in the consensus group 25% of the time (i.e. 10 out of every 40 minutes approximately). It will be worthwhile for users to establish an account with this large miner and send an offending transaction when that miner is in the current consensus group (and send non-offending transactions the other 75% of the time, but note there may be another 10% miner offering the same and so then it is 35% of the time, etc).

So this means the minority miners are starved of income and go bankrupt. This is yet another example of why profitable proof-of-work will in every conceivable design variant always be a power vacuum winner-take-all non-equilibrium.

Sorry profitable proof-of-work can never be stable.