Author: Tier Nolan 2014-05-04 21:11:27
Published on: 2014-05-04T21:11:27+00:00
In an email conversation between Gregory Maxwell and others regarding the failure rates of nodes with random spans of blocks, it was suggested that this approach works out to be asymptotically the same as random for non-error-coded cases. It was noted that if each "block" is actually 512 blocks in sequence, then each "slot" is more likely to be hit, effectively reducing the number of blocks by the minimum run lengths. The conversation also touched on the use of error coding, which can provide an anti-censorship effect where it becomes difficult to provide part of the data without potentially providing all of it. Regarding the current and foreseeable future state of the network, it was noted that there will likely be a reasonable number of nodes that store all the blocks, although perhaps not enough to satisfy historical block demand from the network alone. However, it should be enough to supply blocks that have otherwise gone missing. It was also mentioned that scaling up transactions per second increases the chance of data loss, and with side/tree chains, the odds of data loss in less important chains increases, though they are lower value chains by definition.
Updated on: 2023-06-08T19:15:34.928275+00:00