Published on: 2020-02-12T22:50:58+00:00
The discussion revolves around the Ant Routing proposal for Lightning Network. The proposal aims to obfuscate the counter to avoid giving information to immediate neighbors about the origin or end of the transaction in a highly connected network. The assumption is made that the network has thousands of nodes and node connection larger than ten, where triangulation is not feasible. However, there may be sections of the network that are more "plane"-like than the assumed infinite number of dimensions, meaning a limited number of nodes can be used to usefully triangulate.The proposal solves the problem of payment failures during the pheromone phase by forwarding the pheromone to a neighbor only if the channel balance allows the amount of the transaction to go through. However, the possibility of an intermediate node failing after propagating a pheromone but before it could be routed through must also be considered. Additionally, if the payer has only one channel, intermediate nodes need to store other pheromones with higher distance counters as well.It is important to consider the assumption that "the payee must be online to receive" which should be upgraded to "the payee is *contactable by the payer* to receive". This is because the payee may be online but not contactable due to various reasons such as being on an onion service. Any routing proposal that requires continuous communication between payer and payee makes this unreasonable assumption.Finally, the proposal suggests using just a one-byte distance counter, assuming that the network graph cannot have a diameter greater than 127 hops. With wraparound, a less than operation based on subtraction modulo 256 can be implemented to determine which of two distance counters is lesser or greater.The article discusses the proposed alternative to Lightning Network's source routing, Ant Routing. The authors of a new paper on ant routing have responded to comments from ZmnSCPxj, stating that they are assuming a large and highly connected network with a different geometry from the plane. They want to obfuscate counters to avoid giving information to immediate neighbours about the origin or end of the transaction, but they are not considering optimization with a Dijkstra type algorithm. The authors claim the purpose of their paper is to prove scalability up to 10,000 tx/sec, and they don't claim anything else.ZmnSCPxj proposed Ant Routing as a way to reduce the size of the Lightning Network by eliminating the routemap and making every channel unpublished. However, this system has some drawbacks. If a surveillance node can match both payer and payee seed pheromones, it can determine upper bounds on the distance to both nodes, which is a privacy concern. Channel rebalancing is also difficult with Ant Routing since the shortest path to oneself is to do nothing. Additionally, someone could spam the network with pheromones for payments they won't receive, leading to initiator-does-not-pay attacks.The authors disagree with ZmnSCPxj's statement that Alice doesn't know the channel balances of other nodes and says this problem is solved during the pheromone phase; a node forwards the pheromone to a neighbour only if the channel balance allows the amount of the transaction to go through. They believe spamming is unavoidable but nodes are free to relay the seeds they want, and scrutiny from nodes is required. The authors claim that theoretical O-estimates seem irrelevant to them since any O-estimate is dependent on implicit constants and hardware used, and the argument is given in Section 6 of the paper.Ant Routing prevents path decorrelation since there's only one pheromone seed recognizable for the entire path. Further, complicated payment schemes that take advantage of homomorphism may be hindered by the pheromone system. Confirming path length is an additional step, but if a node under-reports the distance a pheromone travels, it can hijack and acquire information about the payment amount and hash/point, compromising privacy.Routing failures are harder to handle with Ant Routing, but payment failures may be reduced since a recent pheromone indicates the forwarding nodes along that path have recently been online, and the chances of it going offline soon are low.The article also mentions that surveillors can easily determine payments and the maximum distance to the destination and likely source with Ant Routing, which is the same as current Lightning. However, there is already a proposal to remove this issue in Lightning through the use of payment points, but it may not be usable with Ant Routing.Despite the drawbacks, the author notes that constant factors are much lower with Ant Routing, and it may remain practical if certain optimizations are made. The main objection against Ant Routing is that it eliminates the source-pathfinding onion-routing scheme, which reduces how much data needs to be propagated across the network, thus making it more scalable.The authors of a new paper on ant routing have responded to comments regarding intermediary nodes, spamming, payment failures, and scalability.
Updated on: 2023-07-31T22:40:37.684364+00:00