Routemap scaling (was: Just in Time Routing (JIT-Routing) and a channel rebalancing heuristic as an add on for improved routing success in BOLT 1.0) [combined summary]



Individual post summaries: Click here to read the original discussion on the lightning-dev mailing list

Published on: 2019-04-09T10:32:00+00:00


Summary:

The discussion on the Lightning Network (LN) revolves around various proposals and concerns aimed at optimizing the network's efficiency, security, and scalability. One proposal suggests using a global quadtree structure to organize nodes, with each bucket representing a view over its four child nodes. This approach aims to improve efficiency by reducing route length and optimizing for reduced querying rather than finding optimal routes. Anonymity is emphasized by not revealing how much each node knows about the network, and lookups can be delegated to nodes more likely to know the destination.In an email exchange, ZmnSCPxj proposes a global tree with buckets and child nodes, while m.a.holden raises concerns about the potential impact of taking down a node near the root. ZmnSCPxj argues that his approach does not have a hierarchical structure, with all nodes being equally important. The Lightning Network's gossip protocol is discussed in relation to how nodes filter incoming information. Nodes probabilistically filter or signal peers to do so based on given filters, but concerns about DDoS attacks are addressed by failing fast instead of admitting ignorance and failing queries outside of a specified bucket.Mitigation strategies for DDoS attacks include detecting targeting and setting a threshold for expected savings from spilling. The proposed quadtree approach involves a global tree with buckets and child nodes. A distance measurement strategy maps the LN universe onto a circle, making it harder to disrupt through targeted attacks. Probabilistic approaches for knowing nodes near oneself are suggested, along with other gossip strategies.The writer proposes changes to the Lightning Network to avoid targeted attacks and disruptions. They suggest that nodes do not reveal their knowledge of the network and delegate to other nodes with better information. The use of a distance measurement to map the network onto a circle is proposed to make it harder to disrupt through targeted attacks. The author also proposes a quadtree-based solution for reducing route length and improving efficiency. This approach ensures that all nodes remain equal participants and optimizes for reduced querying rather than finding optimal routes. Autopilot suggestions prioritize local channels, but trade-offs may be involved.Discussions also revolve around achieving determinism and diverse distribution of network maps. One proposal suggests using a quadtree based on node ids or hashes, while others are hesitant about hierarchical designs and suggest an artificial "distance" measure based on hashing a node's public key. Pruning heuristics and querying for specific depths are also discussed.The conversation delves into the pruning heuristic in the Lightning Network, which favors high-capacity channels over low-capacity channels. However, this heuristic is gameable and incentive-compatible for nodes with limited storage. Suggestions are made to replace aggregate channels with high-capacity channels and implement JIT routing. Spontaneous payments and complete knowledge of the network are also addressed.Overall, the discussions aim to optimize the Lightning Network through quadtree organization, anonymity, determinism, and efficient routing. Various proposals and concerns are raised to improve the efficiency, security, and scalability of the network.


Updated on: 2023-07-31T21:33:45.715979+00:00