The Problem with Byzantine generals__byzantine

There are many different blockchain consensus algorithms being promoted. Academics like to evaluate this algorithms to the if they solve the Byzantine generals. This particular problem are how to get a group of lieutenants to agree on whether to retreat or advance when up to⅓of the M May is dishonest. In terms of blockchains, the challenge are to get the entire world to agree on a particular blockchain.

One of the aspects of Byzantine fault tolerance (BFT) is, that, algorithm always reaches a conclusion and never ends up I n an indeterminate state. In the other words, at the some point people need to know with certainty, "everyone is in agreement and" a minority of bad Actors cannot halt the process completely. In all systems an assumption that⅔of the generals are honest is a mathematical requirement. Proof of Work does not Solve Byzantine generals

Bitcoin and other Proof of Work algorithms define the best chain as the one with the most difficult Proof of Work. A problem with proof of work, from Byzantine perspective, are, it is always possible for the blockchain, It just becomes increasingly unlikely and uneconomical. In order to consider Bitcoin and similar blockchains to have BFT you must assume this 99.999% certainty is the same as 100 %. Someone with 10% of the hashed power can pull off a double spend against Someone waiting for 6 confirmations once per month Or 0.02% of the time. Cosmos White Paper

The Cosmos White Paper claims a working Byzantine fault tolerance algorithm whereby a group of blocks producers can Unanimo usly agree on the next block. They claim their process guarantees the network to advance as long as more than⅔are honest. The rest of the world never has to consider blockchain forks because they are "impossible" under their.

I was thoroughly impressed at the thoroughness of their approach and the robustness of their. I was also very appreciative this some of their team took the time to evaluate my own algorithm, delegated Proof of Stake (DPOS). Delegated Proof of Stake (dpos)

Dpos is a algorithm whereby stakeholders elect a odd number of block producers. Each round the blocks producers are shuffled and assigned a time slot in which they should produce. The longest chain is considered the best chain. It is impossible for a minority of blocks producers to create a longer chain because they would be producing fewer p ER minute than the majority blockchain.

Like Bitcoin, under this model alone your never have 100% certainty because the majority could switch to a suddenly Y Fork and stop producing on their own. Eventually the minority fork would become longer and global consensus would necessarily change.

We then introduced a new concept known as the last irreversible block (LIB). This is a blocks which has been confirmed by⅔or more of the elected block producers. No node would automatically switch to a fork that isn ' t built in top of the LIB.

Members of Cosmos community recently discovered a particular sequence of events this could cause the LIB algorithm to B Reak down by causing the network to split on two different LIB. This is vulnerability in the LIB algorithm does is not constitute a breakdown of Dpos because the LIB algorithm is an optional Addition designed to identify the number of confirmations required before a exchange has sufficient guarantees that a blo CK won't get orphaned. The LIB algorithm is similar to the 6 confirmation algorithm used by Bitcoin. It works 99.9999% to the time. In over 2 years of operation and millions of blocks it has not yet experienced the kind of deadlock condition is theoretically possible. the real problem-who are the generals?

A problem with Cosmos was an in implicit assumption this set of generals is fixed. At least to a given block. If at any time over⅓of the generals goes rogue, then the network halts and there are no process for resuming the network.

Under Dpos, if a large minority of generals goes bad then the network still the operates to "pending" state. Blocks are produced, votes are cast, and new generals can be elected. Eventually the stakeholders can reach a point where 100% of the generals are honest and the network resumes normal Operati On. In theory, if generals fail the remaining 5% can still hold a election to replace them.

From-we can conclude that Cosmos has a different deadlock condition. Dpos is robust due to it ability to continue operating on a reduced capacity/confidence level even when the majority of The generals are down. Trade Offs

Every algorithm makes certain tradeoffs. Cosmos has 100% guarantee of no forks and can produce blocks every 5 to ten seconds on a distributed network. To achieve this with the same number of generals as Steem, they consume network bandwidth similar to a sustained 5 tr Ansactions per second.

On the other hand, Steem rarely has any missed blocks and even more rarely a has block. Statistically speaking, a block produced while there is 100% participation of generals (producers) has a 99.9999% chance O F being the final block. After two blocks (6 seconds) The probability being orphaned falls. Anyone relying on the last irreversible blocks algorithm (seconds) is protected against almost every conceivable scenari O with a worst-case (purely theoretical) being stuck to a pending state on a minority fork until manual. This means is the IT fails "safe", in a the where you don't accept a bogus transactions.

There is one additional level of confirmation that could to implemented in Dpos:require 100%. With the seconds of confirmation and 100% participation then you can be certain this you'll never get "stuck" and "E" Very transaction you are truly irreversible and globally accepted.

What you should notice with this numbers is this extra time buys to the last insurance of 0.00001% Byzantine. This is a level of the confidence that takes far more than 6 confirmations into Bitcoin to reach, especially with an attacker th At has 33% of the hashed power. Conclusion

Due to the diminishing returns (economic marginal utility), the value of increased confidence from 99.9999% to 100% Is unmeasurable for the vast majority to real world transactions. People lose more from transaction fees than they would if 1 out of every 1 million transactions were to be randomly revers Ed. Meanwhile, the cost of gaining this 0.0001% of confidence is significant network overhead, slower D Complete Failure If more than⅓of the nodes go down.

Both algorithms have their place, but I am very happy and the design tradeoffs made for Dpos THMs available.

https://steemit.com/blockchain/@dantheman/the-problem-with-byzantine-generals