roadmap

Roadmap & status

Status

Phase	Milestones	Status
1 — RVZ reader	container parsing, tables, codecs, packing, Wii partition rebuild, CLI info/decode	✅ done
2 — Legacy decoders	blob abstraction + magic detection; WIA, GCZ, CISO/WBI, WBFS, TGC, NFS	✅ done
3 — RVZ writer	RvzWriter, encoders, junk packing, CLI convert	✅ done
4 — Distribution	NuGet package (net8.0/9.0/10.0), multi-target tests, progress/cancellation API	✅ done
5 — Reference alignment	audit against dolphin-master + rvz-1.0.3; every finding fixed or documented	✅ done
6 — Real-world validation	276 real-file tests (RVZSharp.Slow.Tests) against 90 GameCube/Wii RVZ games (No-Intro SHA-1) incl. writer round-trips — found & fixed the 2 MiB ticket-key writer bug	✅ done
7 — Writer performance	parallel group compression (MaxThreads / --threads, Dolphin's worker-pool model), convert --verify hash comparison	✅ done
8 — GCZ writer	GczWriter / convert -f gcz: 16 KiB zlib blocks, raw-block fallback, per-block Adler-32, parallel block deflate	✅ done
9 — Filesystem & concurrency	DiscFileSystem FST parser + PartitionReader, CLI extract (list/extract/single/partition/gameonly, system data), parallel full-image decode (--threads), async API (CopyToAsync/ReadFullyAsync/WriteAsync), -c purge for WIA	✅ done
10 — API ergonomics	Dolphin-default compression level 5, DiscInfo metadata API, split plain ISO (.part0.iso), path overloads (RvzWriter.Write/RvzReader.Open), seekable BlobStream, Scrub writer option	✅ done
11 — Robustness & performance	thread-safe ReadAt with a 16 MiB LRU decoded-unit cache, binary-search area lookup, ArrayPool group reads; parser caps + mutation/fuzz robustness suite; Native AOT/trimming validated (library annotated, CLI trimmed + AOT publish smoke-tested)	✅ done
12 — Formats & verification	CisoWriter/WbfsWriter/TgcWriter (+ convert -f ciso|wbfs|tgc); DiscVerifier Wii h0/h1/h2/h3 + TMD/H3 verification (verify --partitions); env-var real legacy-file tests + optional dolphin-tool/wit/wwt differential tests	✅ done
13 — Tooling & packaging	GitHub Actions CI (build + fast tests on net8.0/9.0/10.0, coverage artifact, pack + API validation, CLI publish) and Dependabot; PackageValidationBaselineVersion (1.0.0) diffing the public API; SPDX SBOM embedded in the nupkg; CLI --json, - stdin/stdout and shell completions; per-tag GitHub Releases with CLI bundles + NuGet push; docs/ mirrored to the wiki and deployed to Pages	✅ done (1.1.0 published 2026-10-09)
14 — Correctness review	commit-by-commit audit before 1.1.0: retail ticket-key decryption with the Wii common key (WiiVolume.GetTitleKey), extract TMD/cert/H3 partition-relative offsets + H3 guard, WBFS header size/copy fixes, convert --json stdout hygiene, Ctrl+C exit 130, monotonic writer progress, original-exception propagation from parallel loops — each pinned by a regression test	✅ done
15 — Audit follow-up	second commit-by-commit audit: zero-group decode of all-zero partition chunks, extract path-traversal guards (-s/image names cannot escape -o), FST validation hardening, decode -h, help-run telemetry/update-prompt skips, stdin-spool cancellation in every command, release/wiki workflow fixes — each pinned by a regression test	✅ done

Supported

• Read: RVZ, WIA, GCZ, CISO/WBI, WBFS, TGC, NFS, plain ISO — auto-detected, random-access.
• Write: RVZ, WIA, GCZ, CISO/WBI, WBFS and TGC from any of the above (None/Zstd/Bzip2/LZMA1/ LZMA2 with Dolphin's level rules incl. negative Zstd "fast" levels; PURGE for WIA; chunk sizes 32 KiB–2 MiB powers of two or multiples of 2 MiB above that for RVZ, multiples of 2 MiB for WIA; GCZ/CISO with power-of-two block sizes; WBFS clusters ≥ 32 KiB; optional RVZ packing; --sha1-verifiable output). CISO/WBFS store all-zero blocks absent (shrinks scrubbed images); TGC is GameCube-only.
• --scrub: zeroes the data of non-game Wii partitions (update/channel) before converting.
• Wii partition optimization with hash exceptions, FST split, zero groups, PRNG-junk packing with seed recovery; partition keys are common-key decrypted from the ticket (WiiVolume.GetTitleKey) or taken from the container's partition table.
• Progress/cancellation on both encode (RvzWriter.Write/WiaWriter.Write) and decode (IBlobReader.CopyTo/ReadFully); DiscHasher computes CRC-32/MD5/SHA-1 in one pass; DiscVerifier walks Wii partition h0/h1/h2/h3 hash trees and the TMD/H3 tables (Dolphin's VolumeVerifier), reporting per-partition issues.
• Parallel group compression in the writer (MaxThreads, CLI --threads; output is byte-identical to sequential), parallel full-image decode for RVZ/WIA, convert --verify (input vs output hashes), and an async API (CopyToAsync, ReadFullyAsync, WriteAsync).
• File system access: DiscFileSystem parses the GameCube/Wii FST (case-insensitive lookup, file streaming) and the CLI extract command lists/extracts the tree plus the standard system data per partition (DolphinTool-compatible layout).
• Real-world validation: 90 real GC/Wii RVZ images decode byte-for-byte to their official No-Intro SHA-1s; real images re-encode to RVZ (default 2 MiB chunks) and decode back to the same hash. See testing.md for the suite details.

Known limitations

Limitation	Detail
WBFS conversion is slow	WBFS reports a fixed 9.4 GiB logical image; converting reads all of it (mostly zero clusters). decode + convert on the ISO is faster in practice.
PURGE output	PURGE is WIA-only; WiaWriter and the CLI (-c purge) support it, but RVZ readers reject PURGE containers.
NFS key location	the AES key must come from code/htk.bin next to the content/hif_000000.nfs file (or be supplied via the library API).
Sequential random access	full-image decode can use a worker pool (RVZ/WIA, --threads), but individual ReadAt calls and the other formats decode sequentially.

Open questions

1. Real-file validation — ✅ resolved for RVZ: 90 real GameCube/Wii games decode byte-for-byte to their official No-Intro SHA-1s, and real images re-encode to RVZ and decode back. Legacy-format real files (GCZ/CISO/WBFS/TGC/NFS/WIA) are covered by env-var-driven tests in the slow suite (RVZ_REAL_GCZ etc.) plus optional dolphin-tool/wit/wwt differential tests; running them still needs the real files and tools on the machine.
2. Performance targets — ✅ resolved for writing and full-image reads: group packing/compression and RVZ/WIA chunk decoding both run on worker pools (MaxThreads); per-range ReadAt remains sequential by design.

Possible next steps

• Run the legacy/differential slow tests against a real collection of GCZ/CISO/WBFS/TGC/NFS files and dolphin-tool/wit/wwt on a machine that has them.
• Keep the 1.1.x line current: dependency bumps via Dependabot and patch releases from master when a regression is reported.