Author: Matt Corallo 2016-05-11 01:12:32
Published on: 2016-05-11T01:12:32+00:00
The email conversation starts with Rusty Russell proposing to use variable-length bit encodings for transactions in Bitcoin, which would result in a smaller size per transaction. Gregory Maxwell raises questions on the reliability and failure rate of such an approach. Russell responds by saying that it could be used as a best-case scenario in which mempool set reconciliation is applied, and if there are any issues, a later more sophisticated approach could be taken. He also suggests that using variable-length IDs would make the process easier. The conversation then shifts towards the idea of avoiding nonces to save time and instead defining an ID. Maxwell points out that this would greatly increase the cost of a collision and affect overall propagation. Russell argues that the impact wouldn't be significant and explains that the recipient would need to encode all the mempool.Peter R suggests enhancing the receiver to brute force collisions, but Maxwell argues that the narrow band between "no collisions" and "infeasible many" could be avoided by adding a small amount of space to the IDs. Russell agrees to add extra bits on the sender's side but leaves the door open for someone else to implement brute force in the future.Matt, who is working on UDP stuff, chimes in to say that the ultimate goal with UDP is not to provide reliable transport and that his desire for diversity in network protocols is to enable different tradeoffs in reliability/privacy/etc.
Updated on: 2023-05-19T23:21:14.706656+00:00