understanding saito

On Understanding Saito's Proof of Routing Work Consensus Mechanism

Much like Christopher Nolan's film about nested dreams, Inception, from which Saito's namesake is inspired, wrapping one's head around the motivation behind Saito's many differences to standard Proof of Work/Stake (PoW & PoS) blockchain consensus can be a difficult intellectual exercise. In the same way that understanding more deeply someone's waking life will give you better insight into their dreams, understanding PoW & PoS more completely than commonly taught will make clear why Saito is designed the way it is.

In PoW/PoS systems the incentive dictates that if you can author a new, valid, block for the chain, and the rest of the network agrees to publish it, you get a reward. In order for an authored block to be considered valid by the rest of the network, the transaction data must be accurate, and the authored block must contain proof that a certain challenge was met, either by giving the result to a difficult number crunching problem (proving work), or verifying proof that the author of the new block won a lottery with their own tokens in the system increasing their odds (proving stake). There is mining/staking to produce blocks, and there are "nodes" responsible for sharing, receiving and judging information from new blocks and transactions.

And this is commonly how far classical blockchains are considered in a macro sense; what's missing from this description are specifics on how the nodes, rather than the miners/stakers might behave. After all, miners/stakers cannot earn rewards without first receiving information about transactions and sharing information about their proposed blocks. But are there good reasons why miners/stakers should always share all their transaction data, or propagate blocks which reward a competing miner/staker, or even to quickly share blocks which reward themselves? The answer is that since these network activities related to sharing information are not explicitly incentivized, strategic miners/stakers with certain advantages will abuse their ownership of this information in ways that harm the openness and decentralization of the network, but increase their rewards.

The first issue this causes is expressed by "The Free Rider Problem" of economics. If a profit driven miner/staker can always get new block date and transaction information from a generous node, it may decide that running a node itself is not necessary, depending on how costly that is. Currently in Bitcoin, the cost of running a node is not very prohibitive, but in chains which need to store more data in the interests of increased functionality (smart contracts, storage, etc.), increased data throughput (transactions per second), or simply because they have a long history, the costs of being a node start becoming appearant. That data which must be stored somewhere for the blockchain to function becomes an ever increasing burden, and rational actors maximizing profits are not interested in hosting all that data for free.

[double check infura numbers and add bit about aws and sources]
Ethereum's network already reveals what happens to a late stage network when node hosting costs become prohibitive and the network fails to subsidize it - because hosting nodes is expensive and not directly incentivized, the miners on Ethereum raking in most of the rewards do not host public facing nodes.Today a centralized organization, Infura, founded by Ethereum co-founders hosts over 80% of Ethereum nodes used for its transaction data. The only way this is profitable is by controlling data on the network or by charging extra fees on transactions for going through Infura (this is where a portion of your Metamask fees go). Either way, the fact a single entity acts as a gateway for the majority of Ethereum should be a massive red flag to anyone interested in decentralization. This is not a problem that is solved by proof of stake, which relieves transaction time and fees - in fact, if PoS allows data on the Ethereum Blockchain to flow in more quickly, it exacerbates the issue.

The second oversight which PoW/PoS harbor comes from the games rational miners can play with information to be more competitive. [Bitcoin whitepaper says nodes should be open? Check and add if true] A simple one is hoarding transaction information, keeping lucrative transactions to yourself so others can't earn the fees associated. The data may still reach other miners, but if it reaches less miners and takes longer its more opportunity for the hoarding party to earn the fees as compared to someone else.

A more insidious tactic, called "selfish mining", involves groups of miners conspiring together to both get a head-start on mining a fresh block and undermine the rest of the network's work so far on what was thought would be the next block. 1(https://www.investopedia.com/terms/s/selfish-mining.asp, https://www.cs.cornell.edu/~ie53/publications/btcProcFC.pdf) These tactics are just part of the game of maximizing profits in Proof of Work derived incentive mechanisms, and although some Proof of Stake systems attempt to explicitly solve these issues, they do not address the underlying issues regarding what the network has incentive to reward, and while more complicated, simply end up shuffling the problem around rather than outright addressing it.

These exploits in the system allow those with the power to abuse them, usually coming from the collusion of multiple large groups pooling their power in the system together, will not outright destroy it, rather they can use their power to earn more fees than they otherwise might, or other hash/stake power would earn (proportionally) - the most effective groups would be the most centralized, and it would be hard to notice unless they decided to attempt a high profile transaction reversal or one went investigating into how the nodes behaved. [Richard Parris has a blog post which could be linked here]

So the "waking life" of classical PoW/PoS blockchains is that mining/staking is given resources explicitly by the system, but sharing of transaction data and up to date blocks is not, and hosting the data on the network is not, meaning that in order for nodes to make money, they must either act in service of a miner/staker owned by the same party (and with no incentive to share outside that party), or monetize their services which requires centralization like Infura. Much like Inception, Saito is not interested in playing whack-a-mole addressing all of these issue separately in the top layer; instead it delves into the most fundamental aspects of the system to make changes which flourish into a system which explicitly funds the entire network infrastructure solves these issues from the incentive layer.

One big change this brings to the practical functioning of Saito is that efficiency is rewarded by the system. This means high data throughput which does not compromise security - in fact, the more data flowing through the network, the more expensive it becomes to attack. To understand, consider a high level overview of how a new block is created, then added to Saito:

The first qualification to produce a new block is that the node proposing it must have accumulated a certain amount of "work" to be considered valid. Think of the work needed as a score. A proposed block increases its score by accepting transaction data from other nodes, or straight from an end user composing a transaction. The routing work is calculated based on the fee included in the transaction - the first node to receive the fee gets work equal to the fee, and secondary nodes receive one half of what the first node received in terms of work, and on and on with work received being cut in half each time its shared.

It's helpful to skip ahead for a second and say that the rewards a node receives for its work will on average and across time will be proportional to the amount of work it contributed to whichever block gets accepted. Nodes have an incentive to accept transactions to increase their "work score," but they also have an incentive to share that same transaction data in order to increase the odds that the new block contains transactions they helped route, since their work doesn't count if it wasn't included in the block. All of the sudden, spreading transactions as far and wide as quickly as possible is the work which subsidizes the network.

So nodes acting strategically to maximize work will optimize for collecting transaction data straight from users. But even if a node cannot source a large volume of lucrative fees straight from users, if they can collect transactions and fees from nodes which are broadcasting many high value transactions, even though they only receive half, they could still easily accumulate those fees simply by accepting and sharing them. Despite the fact that some nodes may be in situations where they are too far behind to win the race of accumulating enough work to author the next block, the network still incentivizes them to collect the work as it increases their chances of receiving the block rewards in proportion to that work.