One thing we can’t complain about this year is the pace of change in macOS. So much for anyone who thought that Apple’s greater commercial success with iOS and its consumer devices would result in macOS being put out to grass.
Twelve months ago most of us were still getting to grips with APFS everywhere, except of course for Time Machine backups, clicking our way through privacy protection, and wondering what was going on with notarization. No sooner were we getting the hang of all that than Catalina was upon us, and we’re still gasping for breath and wondering why we’ve ended up with volumes named Macintosh HD – Data – Data – Data.
Mojave didn’t end as sweetly as it should have. Maybe too many engineers in Cupertino were working flat out on its successor, but later updates to macOS 10.14 didn’t fix the remaining bugs which prevented it from being one of the better major releases of macOS. Problems persisted if not worsened with firmware, so much so that Apple has had to continue trying to fix them long after 10.14.6 should have been rock solid and a finished piece of work.
Catalina was always going to be traumatic for many, even before WWDC revealed its full potential for disaster. The removal of support for 32-bit software alone was guaranteed to make many users bitterly unhappy, and everyone from Apple, through magazines such as MacFormat and Mac|Life, down to singlehanded blogs like this one, kept explaining what was going to happen if you upgraded in the autumn/fall. I still see comments and complaints from users who seem to have escaped this prolonged barrage of warnings, and are now acting all surprised.
What had been generally underestimated, though, was the impact of splitting the normal boot volume into a Volume Group. As with the conversion to APFS, Apple’s engineers did a remarkable job with a conversion process which worked seamlessly for the great majority of upgrades. But Finder’s fine illusion that nothing has really changed is a thin veneer over much deeper troubles. Too many users are ending up with multiple Data volumes, and getting clean re-install to work properly requires a two-day training course – another opportunity for Apple’s in-store classes perhaps.
Catalina’s new volume architecture also required substantial changes to Time Machine, which are impressive, but weren’t even mentioned in any documentation from Apple. The usual defence of Time Machine being transparent to the user and of no concern to the developer has failed, in the many previously reliable backups which have had to be deleted and started from scratch after the upgrade. This is particularly critical, as once those old backups have gone, many users have no route back to Mojave and may be trapped limping along in a partially-functional Catalina. If they then try to perform a clean re-install, they’re likely to end up with multiple Data volumes, and suddenly even Windows starts looking like an attractive way out.
Other undocumented internal changes in Catalina are also concerning. It’s not that long ago that it looked as if Mac OS X was going to abandon resource forks and make all files single-forked, and its longstanding deprecation of framework support for resource forks seemed to discourage the use of their successors, extended attributes. Yet macOS now makes more extensive, even profligate, use of extended attributes than ever before. Quarantine flags are written to file system metadata whenever a sandboxed app opens a document, and in Catalina a brand new extended attribute is used to maintain access to individual documents under privacy protection.
For those using SSDs, this proliferation of extended attributes in file system metadata can result in some bizarre effects which by anyone’s standards are bugs, but have in Catalina become standard features. For those unfortunate enough to have booted from a hard disk (now required to be in APFS), it’s another substantial performance hit, as their drive has to seek heavily fragmented metadata and the document’s stored data fork.
Catalina’s detractors like to crow about its extended privacy protection, but I suspect for most users these complaints are overdone and unreal. For journalists who might be installing several apps a day, there will be a seemingly endless succession of trips to the Privacy tab in Security & Privacy. For the great majority of users, though, negotiating the new protections in macOS 10.15 isn’t onerous, it just causes us to ignore details in the dialogs – just as we now apparently do over the first run of notarized apps.
Unlike most other changes in Mojave and Catalina, 18 months after its introduction, notarization still isn’t fully in force, and won’t be until February 2020. What seemed to be relatively straightforward requirements when they were first announced at WWDC 2018 have had to be relaxed in order to let some major third-party developers comply. Apple itself has blown hot and cold: in Mojave, first launch of any quarantined notarized app was heralded by the proud warranty that Apple had checked that software for malware and didn’t find any, so assures its wholesomeness. When this was removed in Catalina, it speaks volumes that hardly anyone noticed for several months, and that Apple hasn’t explained why it no longer provides that explicit assurance.
More worrying, though, has been the succession of apps which have been evicted from the Mac App Store not because they failed any of Apple’s checks, but because users and security researchers discovered that they were stealing or leaking user data. Apple’s commercial interests seem not to coincide with the goals forcefully expressed by its engineers when describing the next round of security and privacy protection at WWDC each year.
In engineering terms, 2019 has been a year of great success for Apple, with its problems largely stemming from unintended consequences which it now needs to address, and its staggering failure to communicate in documentation.
A good example of this is the support article explaining How to erase a disk for Mac, published quietly almost two months after the release of Catalina. It describes a major change to a fundamental process which all Mac users need to know, but is incomplete and for many users doesn’t work. There seems little point in creating impressive illusions in the Finder when users can’t rely on your explanation of how to erase a disk properly.