Tomorrow, 5 June, Apple opens its annual developer conference, WWDC. As usual, forecasts of new product releases have made good headlines, no doubt proving effective clickbait, but startlingly shallow. Siri speakers and a new iPad Pro might seem significant, but it’s what is happening inside macOS 10.13 and iOS 11 which are crucial to Apple’s future.
Whatever else it brings, macOS 10.13 will be the most major change in Mac operating systems since Mac OS X itself. Apple is expected to confirm that its new file system, APFS, will be a part of that release this autumn. The last time that Apple made any substantial change to its primary file system was in Mac OS 8.1, back in January 1998, when it released the current Mac Extended or HFS+ file system.
APFS has already been running on all iOS devices which were updated to 10.3, which can give us some confidence that its use in macOS is not going to be a complete disaster. But it is a far cry from the limited storage and walled garden of iOS to the bewildering variety of storage systems and software jungle of macOS.
The first of the key questions will be whether macOS 10.13 will require to be started up from an APFS volume. There is no good answer to this: if it does, then that will slow adoption of 10.13 as many cautious users wait and see how it works out before committing; if it doesn’t, then adoption of 10.13 may be more rapid, but many users will probably stick with HFS+ for the time being.
The next key question is which variant of APFS will be standard for macOS. Like HFS+, APFS comes in two variants which differ in their case sensitivity. Unfortunately, Apple has made a design decision that APFS itself treats folder and file names as just collections of bytes. I have already looked in detail at the consequences of that with respect to the case-sensitive and case-insensitive variants, and neither retains the simplicity of HFS+.
The smart money says that macOS will standardise on the case-insensitive variant, which for now at least presents the fewest problems to the developer, administrator, and user. But we should expect some problems with that, particularly when using non-English text and filenames. These could limit adoption of macOS 10.13 outside the English-speaking world.
Two major benefits which come in APFS are its built-in support for encryption at different levels, and its ability to make snapshots. The former will undoubtedly be used in FileVault 3 and more limited encryption services, which can only be for the good.
Snapshots should transform the way that we make backups, replacing Time Machine in its entirety. Except that we cannot simply replace Time Machine at a stroke. First, there’s the problem of what happens to existing backup systems based on Time Machine. These work using the hard links in HFS+, which are unlikely to be supported in APFS.
My current Time Machine backup is a little over 3 TB in size, stored on an external hardware RAID system, and dates back five years to 2012. Are we going to have to abandon those old backups in favour of new snapshots?
Time Machine is still going to be needed for backup systems which cannot switch to APFS, and so cannot accommodate its snapshots. Anyone who currently performs backup to a NAS, including Apple’s Time Capsules, is going to have to continue with a traditional backup method such as Time Machine at least for the time being. Eventually, we’ll all want to make snapshots to APFS, but Apple needs to provide a smooth transition to that for a very large number of Mac users.
Finally, and most important of all, macOS 10.13 is going to have to offer extremely reliable in-place migration from HFS+ to APFS for startup volumes. No one is going to risk upgrading to 10.13 if they have to convert their startup volume to APFS, and that process is not bombproof. Even tiny error rates would rapidly propagate over social media to make the 10.13 upgrade a near-instant catastrophe.
Linked with that, of course, is the issue of access to existing Time Machine backups. If the only safe way to upgrade to 10.13 is with both those and a complete mirror copy of your startup volume, Apple is going to have a very hard time persuading users to upgrade.
There’s a lot more that we need to know about macOS 10.13, such as whether it is built from the kernel up to tackle tomorrow’s security challenges, or whether Apple is just going to keep applying more duct tape and hope that the threats go away.
Some would want Apple to abandon its new unified log system, which it introduced with Sierra. For my part I am quite happy for it to continue, but only if Apple provides much more powerful tools for working with the log. I cannot believe that Apple’s engineers are stuck with the emasculated Console; if that is insufficient for them, so it is woefully inadequate for the rest of us.
One thing that I am confident of, though, is that Apple will not announce any campaign to improve its developer (or user) documentation, nor will it tackle the steady accumulation of irritating or embarrassing bugs. Let’s at least hope that it gets the big things right. Among those, of course, are some new desktop systems, such as an iMac which uses a modern chipset.
We’ll soon see.