In order to function properly, the DPOS process depends on the cooperation of token holders to cast votes for the block producers they are electing to act on their behalf.
This is not relevant to BFT. Ability to transfer/send funds directly or indirectly through leasing/delegation between those producing blocks / nodes / pool operators on any PoS blockchain has 0 relevance to BFT.
DPoS has all the guarantees of virtually every blockchain design ever, voting is just 1 more & including many incentives your report doesn’t mention at all.
If voting breaks BFT, so would sending stake.
Also, voting/sending always works as long as less than 1/3 are byzantine. As those are typical boundary conditions for proving BFT, they should be sufficient.
If EOS isn’t BFT, nothing is.
Design: https://medium.com/eosio/dpos-bft-pipelined-byzantine-fault-tolerance-8a0634a270ba
According to the aforementioned description, EOS token holders should be considered nodes within the network.
No, they are not.
Once the votes are placed, without valid signature from voters changing the vote, block producers alone can’t change the votes. At best they can censor future changes (like on any blockchain), but BFT doesn’t need to rely on voters changing votes & if <1/3 are byzantine, nobody is censoring any votes. Note TaPoS also helps support BFT.
15/21 (33% + 1)
you mean 66% + 1
won’t actually present any further optimizations
you think probability of 1 party being dishonest is same as probability of 6 parties being dishonest at same time?
100% of the network is controlled by 21 block producers
By this logic, 100% of every network is controlled by ~3 mining pool operator nodes on Ethereum and even less in any direct bond based PoS https://i.imgur.com/rhPiMiG.jpg
a bad actor now only needs to collude with 28.24% (6/21) of the 21 block producers in order for its cartel to control the entirety of the network.
so instead of 1, they need 6. and not control since they can’t fake signatures of other BP but simply censor — which doesn’t break finality — which is what BFT is all about.
cartels are possible in EVERY blockchain — so tell me how nothing ever can be BFT.
There is no evidence of cartels being optimum given what’s at stake.
have you considered reading a book on basics of blockchains?
your implications of cartel formations are not analyzed vs any incentives or punishments present to prevent it and appear to be based on nothing but popular slogans you overheard.
your report doesn’t even acknowledge ability to tolerate a single byzantine producer to show an example of just how much you had to try to mislead about even basics of BFT in your report to put ethereum into better light.
By requiring a single node within the network (the token holder) to rely on the good intention of another node (a block producer), these functions now rely on human behavior and a certain degree of good faith
Just like every network relies on its block producers to be honest and with incentives in place to prevent them cheating. You clearly don’t understand that block producers aren’t forced by algorithms or math to be honest, but by incentives, which you do not cover for EOS at all and imply everyone else does it without incentives through what sounds like algebra.
there is no mathematically enforced protocol implemented to ensure that block producers are acting honestly 100% of the time
This is never the case. You are not qualified to comment on blockchains. Ever.
The only thing EOS added on top of existing protocols is ability to delegate stake and some pareto breaking rules — all other mechanisms are still present.
The ability of this system to eliminate the need for trust when communicating across inherently adversarial environments
It’s called trust minimization and not trust elimination for a reason as you always depend on trust on some N number of ideally independent entities. There’s no equation that forces honesty. Game theory is fancy way of saying incentives. Bonds are not the only kind of incentives. Bitcoin doesn’t have bonds.
There is so much wrong in report nobody knows where to even start. For example:
Bitcoin also doesn’t have utxo commitments and countless blockchains didn’t have account/state commitments before and after ethereum. According to your definition were also not deterministic or cryptographically secure — maybe your definition is wrong. Just maybe you can derive state from genesis block. Just maybe.
apparently sybil protection makes something a data streaming service — great comparison.
guess what — public/private key based permissions instead of classic public /private key based ownership is still using cryptography. implying account based designs do not use cryptography is nuts.
