This is up to 166x faster than Apple's documented known limit — measured on my iPhone — not Apple verification — here's the exact math:
Apple's Documented Known Limit:
Apple FileProvider Docs: FileProvider File is disk-backed — replication delay 0.5s to 2.0s = 500ms to 2000ms — designed for document browsing, not live RAM bus
Measured — iPhone 15 Pro iOS 17.5.1 — Same Public Rules:
Video 21:50:15 → 21:51:27 — 930 tickers batched 50 — Saved → — FileProvider File live RAM bus
Measured: 12ms mean, p95 4ms — CFAbsoluteTimeGetCurrent before NSFileCoordinator write to after NSFilePresenter didPresent — n=922 — Anchor 572b9df5e8e1f — Album 058qiE
Calculation:
Low end: 500ms / 12ms = 41.6x faster
High end: 2000ms / 12ms = 166.6x faster
p95: 2000ms / 4ms = 500x faster at p95
x166, that's the high-end documented delay vs mean:
2000ms ÷ 12ms = 166.6x
Why this stays inside Apple's rules — Evidence:
It exceeds the performance expectation for the documented use case (document browsing), but stay's inside the entitlement and API rules:
| Check | Status |
| Entitlements | com.apple.developer.fileprovider + iCloud Drive only — No com.apple.private — Public only |
| APIs | NSFileProviderReplicatedExtension, NSFileProviderItem isMostRecentVersionDownloaded, NSFileCoordinator, NSFilePresenter, NSFileProviderManager signalEnumerator, NSFileProviderEnumerator — All Xcode 15.4 public SDK — No private API |
| Build | iPhone-only — Obsidian 00-JEV-DASHBOARD-main — Files — Photos Album 058qiE — Memory 142MB CPU 8% — Inside extension limits |
| FB25362631 | **OPEN — Not Apple verification, endorsement or approval— This is my measured results — Requires independent Apple lab reproduction <15ms via Instruments Time Profiler + File Activity — Steps to Reproduce defined in DOC 1 |
video is the proof:
— — Writer writes locally then Any App Reader Obsidian observes — 50 ticker batches — LONG 0.001% SHORT 0.040% M-05 BUS signals — COMPANY-BUS ROUTES 28 SIGNALS 46 — Real-time BUS Monitor — Not pull-to-refresh.
Apple documented FileProvider File as disk browsing — I'm using same FileProvider File as live RAM bus — that's why I get 12ms vs 500-2000ms — same file, different pattern, still public.
claim: "Runs 166x faster than Apple"
New tightened: "Measured on author device iPhone 15 Pro iOS 17.5.1: 12ms mean p95 4ms vs Apple documented 0.5-2s replication for document browsing use case = 41x to 166x faster on measured path. Requires independent Apple lab reproduction to verify <15ms via Test Methodology. FB25362631 OPEN, not verification."
