Hyperspace
Perceived Need and Value
- Many see this as most useful on Macs, where internal SSD upgrades are expensive and non‑user‑replaceable.
- Others argue that on machines with cheap multi‑TB NVMe or NAS, reclaiming a few GB (e.g. 1–10 GB) is “spare change” and not worth the cost or risk.
- Some users report modest savings (≈1–2 GB on large home dirs) and decide the benefit doesn’t justify the price; others see double‑digit GB savings and are enthusiastic.
How It Works: APFS Clones vs Links
- Repeated clarifications that this uses APFS copy‑on‑write clones (reflinks), not hardlinks or symlinks.
- From the user’s perspective, all file paths remain and can diverge later; unchanged blocks remain shared.
- Multiple questions about “what happens if I edit one copy” are answered: only the modified parts are written separately; the other clone stays unchanged.
Safety, Risks, and Metadata
- Strong general mistrust of dedup tools: fear of silent corruption and irreversible mistakes; some contributors say they only ever log duplicates, never auto‑modify.
- Hyperspace is described (via docs/podcast) as conservative:
- Skips files if it can’t preserve metadata (timestamps, permissions, extended attributes, resource forks).
- Lets users review proposed changes before committing.
- Nonetheless, several people prefer to wait and see how it behaves “in the wild” before trusting it with important data.
Why the Filesystem Doesn’t Do This Automatically
- Discussion of why APFS doesn’t background‑dedupe duplicates:
- Comparisons to ZFS, Btrfs, XFS, NTFS/Windows Server, ReFS, which can dedupe at block or file level.
- Trade‑offs mentioned: high RAM usage, heavy I/O, complexity of on‑write vs periodic sweeps, and operational horror stories from backup dedupe.
- Linux offers kernel syscalls (e.g. FIDEDUPERANGE) that validate ranges before deduping, making userland tools safer.
Alternatives and DIY Tools
- Multiple free/open‑source options cited: fclones (APFS clones via
cp -c), rmlint (with clone mode on Linux), jdupes, duperemove, bees (Btrfs), czkawka, custom SHA‑based scripts, and a macOS CLI tooldedup. - Debate over algorithms: use file size, partial content, or full cryptographic hashes; some argue SHA‑256 collision risk is practically irrelevant, others still favor byte‑by‑byte confirmation.
Pricing Model and App Store Issues
- Pricing is mixed: some praise the “scan free, pay to reclaim” shareware‑style model (one‑month, one‑year, and lifetime unlocks), others say $10–50 is too high for a storage utility.
- Complaints that App Store hides IAP pricing and that you often must dig to see costs, especially from non‑Apple devices.
- Supporters argue you’re paying not for the basic idea (which many FOSS tools implement) but for careful engineering, safety checks, and a polished macOS GUI.
Platform, Performance, and Closed Source
- Requires macOS 15; users on older systems are disappointed and see this as part of Apple’s “hardware treadmill” and SwiftUI’s reliance on new APIs.
- Side discussion: native SwiftUI/macOS table performance vs HTML/webviews/Electron; several say browsers are now surprisingly faster for large tables.
- Some hesitate to run a closed‑source, full‑disk utility on work machines; others counter that open‑sourcing would both enable free clones and greatly increase support burden.