Track H (Open Pair): JavaScript to Go.
qs-go is a clean-room Go port of the frozen ljharb/qs behavior surface. The submission is designed to be inspected, reproduced, and challenged:
- Frozen upstream JavaScript baseline: 1,045 / 1,045 passing at the recorded source identity.
- Go verification: focused parser/stringifier tests plus 672,321 deterministic differential comparisons in 60 seconds, with zero mismatches and zero oracle or Go errors.
- A larger differential corpus exposed and fixed an ECMAScript-observable integer property-ordering mismatch; the regression is permanently locked.
- Reproducible, raw-retained benchmark evidence reports both gains and regressions. Median cold start was 14.9 ms vs 73.3 ms, peak working set was 35.85 MiB vs 65.91 MiB, stringify medians improved 40.3% and 60.7%, while parse medians regressed 20.0% and 11.6%. The first 384.82 ms Go cold-start sample is retained; with n=40, p99 equals the maximum.
- The 4:40 demo includes a continuous live terminal evidence segment, exact commands, hashes, scope, and limitations.
Important scope: 1,045 / 1,045 is the frozen upstream JavaScript baseline, not a claim that those JavaScript tests ran directly against Go. Go equivalence is supported by the documented Go tests and bounded differential corpus. The ordinary differential wire scope is JSON-compatible dense values. The Windows portable Go toolchain used for the final audit had cgo disabled, so the race detector could not be run; this is disclosed rather than hidden.
Assets:
port_mortem_qsgo_demo_final.mp4- final 4:40 submission video.qs_go_release.zip- reproducible source archive.qs_go_benchmark_evidence_v2.zip- complete benchmark scripts, raw samples, summaries, and environment evidence.- PNG assets - cover, fixed-counterexample evidence, and benchmark evidence.
See README.md, PROVENANCE.md, DECISIONS.md, docs/compatibility.md, docs/clean_build_evidence.md, fuzz/, and benchmarks/ for the full evidence chain.