Author: Bram Cohen 2016-07-15 23:00:57
Published on: 2016-07-15T23:00:57+00:00
In an email exchange, Peter Todd suggests using BLAKE2b for internal hashing due to its efficiency in omitting padding when the data is exactly one block in size. There is also a discussion about the implementation of a root branch block with nodes arranged in fixed positions and terminals having outpointers to other blocks. The author then provides an analysis of how well their implementation would work when everything is stored on a regular spinning hard drive, estimating that even a very underpowered node could keep up with the blockchain if it is full of simple transactions and validating an entire blockchain history might take a few days. However, if some layers are kept in regular memory, validating a whole year of history might only take a few hours. Finally, the author acknowledges that the MMR proposal's data structures are complicated and suggests that a similar analysis specifying storage media and expectations of latency numbers would clarify the reasoning significantly.
Updated on: 2023-06-11T05:44:10.896711+00:00