Media library apps have been doing this kind of stuff forever. An index of the files + metadata allows for a better and more performant experience. But, if an entry in the DB gets pooched, file remains on the drive and is hidden from the user.
Many media library apps actually have a way to repair and or rebuild the library DB if it gets out of sync or corrupted. iTunes straight up put that feature in the menu bar. The Photos app will do it if you launch the app while holding command-option.
Back when iPods were king, how many of us had old music come back to life after a fucked up iTunes library was rebuilt? It’s kind of a similar issue.
Given the rarity of this, it could’ve just been the normal random stuff that happens in computer land. Requests that don’t complete because they were interrupted by a crash, the rare bad block, etc. Or maybe it was just a bug that occasionally reared its head under certain circumstances.
Whatever it was, it wasn’t the first time a piece of software had an index that was messed up and out of sync with the stored files.
As for the iPhone storage thing you mentioned, I don’t know what you’re talking about. There was a IOS 17 bug early on where people mentioned that the OS needed a restart to claw back space from temporary install files and caches.
That said, the corrupted DB we’re talking about appears specific to the photos app. It’s not the file system index. It’s basically a glorified preference file.
E.g. iCloud says it’s using 13.4 GiB to store photos, Settings -> General -> iPhone Storage says I can save 15.5 GiB because they’re backed up on iCloud, and if I use idevicebackup2 to pull everything off the phone, there are 21.7 gigs of photos
I’m wondering if these discrepancies are related to the photo app not actually deleting pictures from the filesystem
Weird. I’d bet money on syncing issues, compression, etc. But who knows, if you have a Mac you can rebuild the library, let that sync, and see what happens.
That’s not how lossless compression works. No data is lost.
For example, if you zip a folder of images, then unzip them, the pictures come out with their original sizes and structure. Zip is lossless.
Let’s use the analogy of a dish sponge.
Let’s pretend you wanted to make a dish sponge smaller. Lossy compression would make the sponge smaller by cutting off parts and throwing them away. Lossless would make it smaller by squish the sponge, and it would return to its normal shape once you stopped squishing it.
Then how would they be training AI on it? If they don’t have it? If it’s on device what’s the problem? Deleting a photo doesn’t wipe the bits to 0, it never has