macOS Sonoma is setting records for update size
eclecticlight.co
eclecticlight.co
This article is saying Sonoma has the smallest updates in years, and that installation time is decreasing.
I sort of get it if this was some unimportant background download being done, but for a game you actually would like to play, you are stuck with "you should have started your computer 2-10-20 hours before actually wanting to play" depending on how fast your "Fast internet" is.
Which is weird because many manufacturers don't restrict flashing a different rom, and even with standard Android you have more access than iPhone gives you.
Manufacturers don't, but app makers do. I know people running on LineageOS that have to keep a second smartphone on stock Android for their banking authenicator app (which fails on Lineage because it's not the manufacturer-supplied ROM and that supposedly means it's been hacked).
I agree though software updating is widely inefficient given how it could be done.
For example, Apple Watch series 3 cannot update without replacing the whole system for at least 3 years. In this case they didn't even sell more storage.
In the end, whatever; it's just Apple nonsense, better be done with them and look for hardware elsewhere than care about all this...
Apart from that, the bigger issue is that this is something that affects many users and the selling point of Apple was always that it should just work, which it clearly doesn't in this case.
This lets the system clear space for large downloads like this. You just have to know which options to enable.
Comcast for example has an 1.2TB per month cap and $10 for each additional 50GB chunk.
With 4k shows on Netflix, online gaming, and 60gb device updates it’s very easy to hit the cap.
According to their app, I'm consistently using 3-4TB per month. I don't know what the breakdown is but my guess is it's mostly streaming video and video game downloads.
Say that update could have been optimised down to say 100MB through shipping only binary diffs instead of complete executables, that's hundreds of petabytes of network traffic that doesn't need to be transferred, which means less link congestion, less packet processing, less data temporarily stored not just on customer machines but in datacenters as well.
Unnecessarily large updates in systems with hundreds of millions of users has a non-negligible environmental impact.
There are so many low hanging enormous fruit to pick that it makes almost no sense to start with the very difficult and insignificant ones.
Just look at the iOS 17 OTA packages for iPhone 14 Pro.
- 16.6.1 -> 17.0: 3.3GB - fair enough, new major version
- 17.0 -> 17.0.1: 452MB - security updates and bugfixes
- 17.0.1 -> 17.0.2: 384MB - security updates and bugfixes
- 17.0.2 -> 17.0.3: 450MB - security updates and bugfixes
- 17.0.3 -> 17.1: 1.35GB - this one is interesting, an update from 17.0.2 to 17.1 is actually about 150MB smaller than from the latest version which uses the same upgrade package as 17.0 GM
- 17.1 -> 17.1.1: 395MB - security updates and bugfixes
- 17.1.1 -> 17.1.2: 414MB - security updates and bugfixes
In total, I've downloaded about 3.44GB of updates for my iPhone since the iOS 17 upgrade in September, and have received WiFi AirDrop support, some security fixes, and better Home app Matter support. Security fixes are in general just a few lines of changed code, and should result in minimal changes to a binary, so if you just diff the new binary with the old one it should come out pretty minimal, yet the patches are several hundred megabytes.
Someone in a different thread suggested Apple were doing partial binary patching, but I don't see it.
Obviously this means you have to maintain 2 update tracks, one optimised flow for all the normies, and one for everyone who has disabled SIP. But the benefits are worth it when you consider all the electricity and bandwidth used by all the excess data you're no longer sending over the network.
Combine this with APFS snapshotting and it should make for a resilient update schema which uses minimal amounts of data.
Or am I completely wrong in my assumptions about binary diff patches?
> Since Big Sur, updates have reduced in total size, and in Ventura were close to or below earlier versions
Apple is shipping smaller updates over time, not larger (any more).
Also, I just checked. Even this record "small" sized update is still an upward trend based on when my Internet connection was originally provisioned.
I guess it’s used in this sense here
Is it said in a neutral sense like yardstick or is it an ideal to aspire to like exemplar or paragon?
You can see here a kilogramme étalon: https://upload.wikimedia.org/wikipedia/commons/0/00/Standard...
and a mètre étalon: https://upload.wikimedia.org/wikipedia/commons/2/22/Mètre-ét...
Edit: for completeness sake, a litre and décalitre (10L) étalon: https://portail.polytechnique.edu/musx/sites/pf_191/files/ac...
They still sell 256GB base models when 1tb is $50.
Games are the worst. If you haven't started one up in a week you need to download an important patch - somehow using 6GB in the process.
I'm pretty sure there is just next to no effort put into keeping the file sizes down.
The amount of updates to what should be more or less finished products is just insane and I can't understand why the industry doesn't see it as a issue... Well, money but still.
But thanks for the (snarky) heads up that there's yet another update (29.4GB this time) to download :)
Sure, that package actually contained 2 complete images for different series, but why does a TV OS need a 1.5GB partition in the first place? (And also, why are they bundling two images for different series)
Rhetorical question, I know the answer, but my point is nobody actually likes the features that cause this problem.
What game was this? No games are pumping out multi GB patches per week unless they're early access on steam, at which point yeah, it's early access.
> I'm pretty sure there is just next to no effort put into keeping the file sizes down.
This is a silly viewpoint, and I've had this out many times on this site. We do care, the demands are just growing and people have unreasonable expectations.
Game sizes have ballooned because expectations are huge.
4k HDR textures are 16x larger than 1k textures. HDR is an extra 25% for the increased bit depth. Textures can be 20x larger than they were 10-12 years ago.
Map sizes are increasing (compare cyberpunk to GTA 5 for example).
Loading screens are heavily frowned upon so compression becomes a lever to adjust.
Audio/VO is highly dynamic and there's absolutely loads of it. Take a look at Overwatch, I play a few hours a week of that game and I still here new voice lines every day.
I worked on shipping smaller patches on AAA games at $PREV_JOB. If you compress everything to high hell and back to be stored on disk, you can do a partial update of it anymore. We tried to be tactical and adjust compression/groupings of what would change frequently and what wouldn't, but sometimes we got it wrong.
Occasionally large things change and force something akin to a full download. We don't want it, you don't want it, but that's where the content comes from.
Then you have the actual logic of patching. If we change a 30MB file, that's part of a 1GB compressed block, then it's a 1GB download. We also need to copy the 1GB we're about to replace somewhere to roll back if you lose connection during the update, plus enough extra space for bookkeeping, extraction, etc.
If you think you can fix this, I'm sure most game studios will hire you with the drop of a hat.
Of course, Apple used binary patching and compression and all that jazz... this was for the "full" version that delivered entire files.
Why oh why does a basic Windows/MacOS desktop use tens of GB to install and use gigs of RAM? When it's sitting at an empty desktop, what is it really doing that's multiple orders of magnitude bigger than Windows 2000 on 64MB RAM?
The trend is linear so, it doesn't look like the updates are actually getting bigger over time, rather is a new paradigm of overall larger updates with Apple Silicon Macs.
It's 845 MB in size.
I can't even begin to imagine what can weigh in at 845 MB for a bug/security release.
I hope Apple can stop faffing about with new whiz bang features for a few iterations and instead focus on working down the ridiculous list of bugs in Sonoma. Somehow they've managed to utterly fuck up application switching, the mind boggles.
Edit since I'm being throttled:
I wholeheartedly agree (and I'm sure I've commented to this effect on HN before): this is probably the nicest Apple laptop I've laid my hands on and nearly the worst version of MacOS I've seen in three decades. I really like the 14" pro form factor. Thankfully I have until January to decide if I want to keep the laptop.
Keyboard focus does not reliably follow you to the new foreground application. I saw this mentioned in the comments in the Ars review but it seems to be worse for me than that guy.
Finder does not reliably display a progress indicator when copying big files, and quick look is as unreliable as ever.
AFP support is completely broken. Apple should just rip that code out instead of letting it rot like that. It's particularly frustrating because Apple dropped support for the version of SMB where Samba "properly" supported unix-y permissions.
SMB support worked for a while in the Finder, but would corrupt metadata from the command line. With 14.1.2 it seems like creating/modifying files on an SMB share from the Finder is as unreliable as the command line.
[1]: https://www.bloomberg.com/news/articles/2023-11-07/apple-del...
I figured out that by turning off File Sharing / Remote Management / etc on the Sonoma machine and turning them back on, it fixed it. Hope this helps.
Except for the whole SMB data corruption thing that's apparently been a problem since Ventura.
I can't even begin to understand what's going on these days, but I'm sure there's an explanation.
Whether or not the system partition is mounted readonly for daily use doesn't have anything to do with the update size.
See the person above you showing that attitude.
Changing a few compiler flags, say to enable a hardened stack, or a compiler update that fixes a bug in such a feature probably could do that.
It could mean lots of tiny changes sprinkled throughout the code that the patching algorithm can’t compress efficiently.