[circularfoyers]

I still don't understand why Apple haven't changed their update model to A/B partitions, so if an update can't boot, it boots the previous partition. Android has been doing this for a while and Fedora Silverblue/IoT does the same thing, both of which I've found really useful in the past. If communicated well, it would elviate the update anxiety many people I know have with their iPhones.

[dwaite]

It will not be able to reliably work with the A partition once B boots the first time - the update process even for many minor releases will trigger local and cloud data upgrades, and there are security reasons to not leave the option to leave older/unpatched versions runnable.

I don't think I've had a software update fail since iOS 7 beta 1 - and that was an issue with updating local data on first boot, so any attempt to revert to the prior OS without a wipe would have been pointless.

[circularfoyers]

I believe the point isn't to let you rollback at will, but it'll automatically rollback if it fails a boot check, so before anything is written to the filesystem. However, it's worth noting that Fedora Silverblue/IoT let's you rollback at will without problem.

[MBCook]

What percentage of iPhones actually require DFU after a bad iOS upgrade?

I’ve had iPhones since the 3GS and never experienced it, neither have anyone I’ve known.

I’m sure it happens, but I suspect it’s really rare.

[yjftsjthsd-h]

> and Fedora Silverblue/IoT does the same thing,

Silverblue uses ostree, which AIUI uses hard links in a single filesystem. This is probably better most of the time, because it means minimal storage overhead, though it means you can't change filesystems over an update and you don't resist filesystem corruption.

[wmf]

How much space should they reserve for iOS 24?

[BonoboIO]

Sell the base 128GB model that will be too small pretty soon and sell the cloud storage. Win win.

[igornadj]

How often does a deployment rollback go smoothly? Almost never. The priority is roll-forward and reduced error scopes.