Evidence

This section keeps public performance language tied to a checked evidence boundary. The goal is not to show the largest number, but to separate what the repository can currently prove from what remains a design ambition.

Start here

• Benchmark methodology explains the evidence rules, dataset scope, and reproduction commands.
• docs/benchmark/README.md is the repository-side narrative for the tracked benchmark chain.
• benchmark/results/err091571-local-supported.json is the machine-readable tracked result.
• benchmark/results/err091571-local-supported.md is the human-readable tracked report.
• Research references place the evidence next to adjacent compressors and papers.

Current evidence

The repository currently tracks one public evidence set built from ENA accession ERR091571 using a reproducible 2,000-record smoke-scale subset. Within that documented run, the observed leaders in the measured local-supported toolset were:

• best compression ratio: bzip2
• best compression speed: bzip2
• best decompression speed: gzip
• current fq-compressor result: not leading on those three dimensions in the documented smoke-scale run

Current boundary

The current evidence is intentionally narrow. It does not yet establish a verified specialized-peer ranking because Spring is still tracked as deferred rather than measured. Readers should treat this section as an audit trail for what the repository can currently prove from that smoke-scale run, not as a blanket best-in-class conclusion.

Continue with

• Whitepaper to see how measured results constrain public claims
• Research for papers, comparators, and evolution notes
• Academy if you need to run the tool rather than debate its public positioning