GitHub project bzip3 pitches faster successor to BZip2
The open-source project reports higher compression ratios and faster performance, but its claims have not been independently verified.
A GitHub project called bzip3 is presenting itself as a faster “spiritual successor” to BZip2, with a focus on compressing text and code.
The project combines an order-0 context-mixing entropy coder with a suffix-array-based Burrows–Wheeler transform. It also uses run-length encoding alongside LZ77-style matching and PPM-style context modelling.
Project documentation reports compression speeds of up to 17 MiB/s and decompression speeds of up to 23 MiB/s per thread on some x64 Linux builds using clang13. It says results can vary significantly depending on the compiler, hardware, operating system and build, with Windows and 32-bit versions potentially slower.
The project includes comparisons with other compression tools and describes a possible use after long-range deduplication with lrzip. The supplied material does not provide independent benchmarking or detailed comparative figures sufficient to verify its claims against BZip2.
Bzip3/libbz3 is licensed under LGPLv3 only. Its documentation also warns that bugs could make compressed data unrecoverable and says users accept the possibility of data loss.
