11 ms·
There are others here who say this is a bug, not a feature, and I agree; and having looked through the open-source code that Apple releases, I have a reasonable
by userbinator 7y ago
There are others here who say this is a bug, not a feature, and I agree; and having looked through the open-source code that Apple releases, I have a reasonable guess of how this happened.
Apple seems to have an internal developer culture that basically does not value backwards-compatibility nor stability, and instead really loves rewriting huge pieces of code all the time. If you look at the OSS that it releases, you'll see interfaces between the various components change greatly between versions and lots of other churn. There are many places which look unimplemented, with only stub functions present. With such a situation, it's not hard to envision how a simple "change the MAC address" function could've gotten left out of some piece during a rewrite, because it otherwise does not affect basic functionality.
I really recommend downloading the OSS from Apple and inspecting it; the pieces it copied from BSDs etc. are relatively stable, but the amount of churn in other pieces is surprising to see.
- oneplane 7y agoI'm not sure about the internal culture, but externally they seem to stress the importance of only using public published interfaces and frameworks as API & ABI layers and not touching anything else as they won't guarantee any stability in private internals. (per their docs iirc)
- LeoNatan25 7y agoEven with the so-called "public" APIs, the way Apple operates, there is no guarantee of any stability, as most iOS developers know. AppKit is much more stable than UIKit, but still nowhere near the stability that companies like Microsoft provide (including changing the name of a product—Windows 9->10—so that existing software does not break!). If you are using features of the frameworks that are outside of the "happy path" for that year's keynote announcements, you can expect terrible bugs on a yearly basis. These bugs happen for the exact same reasons as stated in the GP.
- oneplane 7y agoBut doesn't that go for anything? Take android where you need to account not only for all the cases of versions but also for every differentiated version from every brand and device, or if you build for things like silverlight which gets killed with no replacement. There are always things changing and unless you are on the same path as the vendor you're gonna get screwed. It's always been like that.
- LeoNatan25 7y agoI think you misunderstand what I mean. There is an API, which is documented to function in some way. You'd expect it to continue functioning like that down the road, but since Apple engineering has such poor QA and is always rewriting things, their developers just let things slip through the cracks, often creating broken API for years. Android whataboutism is not really productive here.
- oneplane 7y agoOdd. I never had a public API go bad on macOS, ever. There were always multi-year planned deprecations etc but no sudden changes. The only issue I remember had to do with a toolbox issue before my time where timing wasn't always reliable to be sub-ms precise. Even 68k to PPC transition went fine API-wise, as well as toolbox to carbon. Perhaps you are exclusively referring to mobile 'apps' and the APIs you can use for those?
- LeoNatan25 7y agoYeah, the context of the post you replied to is for iOS APIs. AppKit is a lot more stable, and it shows. It is still possible to run 10-12 year old software on Mojave (but not on Catalina). On the other hand, sometimes even one OS upgrade can break existing software visually or worse on iOS.
- gggmaster 7y agoIf a proprietary software is not much more better than an open source one, I will not choose it.
- cerberusss 7y agoCan you give some examples of those "terrible bugs" you have experienced on a yearly basis? I have developed for iOS for the last 5 years, and I haven't ever experienced those in UIKit. This year brought one single bug in UISegmentedControl, in a situation that doesn't happen often: https://forums.developer.apple.com/thread/122736 https://forums.developer.apple.com/thread/122736
- LeoNatan25 7y agoUINavigationController—https://twitter.com/LeoNatan/status/1168286477882728448 https://twitter.com/LeoNatan/status/1168286477882728448 UITabBarController—https://twitter.com/LeoNatan/status/1168286996915204097 https://twitter.com/LeoNatan/status/1168286996915204097 UIActivityViewController Various UIView layout issues Safe areas—https://twitter.com/LeoNatan/status/1179475940784721920 https://twitter.com/LeoNatan/status/1179475940784721920 UITableViewController—iOS 13's sheet presentation broke clearsSelectionOnViewWillAppear (https://twitter.com/LeoNatan/status/1137474148488364034 https://twitter.com/LeoNatan/status/1137474148488364034), for example, but it's often broken: https://twitter.com/Javi/status/644631570083483648 https://twitter.com/Javi/status/644631570083483648 SLComposeServiceViewController is completely broken on iOS 13
- cerberusss 7y agoYou quoted a bug in 2015, then three in 2019. That's not on a yearly basis. Aren't your views colored by a bad iOS 13 launch? Because I do agree with you that iOS 13 seems bad (not my personal experience, from the general clamoring around us).
- LeoNatan25 7y agoNo. I maintain a pretty popular iOS open source project, as well as many low level projects for my job. It’s always unstable. Every year, new bugs pop up. Is it the end of the world? Probably not, but is it acceptable? It shouldn’t be. Before 10.15, I could run Mac software built 15 years ago. If AppKit developers can preserve backwards compatibility, so should UIKit developers.
- LeoNatan25 7y agoThis is true about all walks of Apple software, sadly, from file syncing to Bonjour (mDNSresponder) to UI frameworks to developer tools to consumer applications. Apple developers rewrite software in a butterfingered way on a consistent basis, creating software that lacks any design, missing large chunks of functionality and often lacks any kind of polish or coherency. Other problems in the dev culture accentuate this problem even further. It seems there is a complete lack of dependency graph between the software stacks; apple engineers who opt to rewrite a piece of software or functionality have no documentation or idea how the software is used or was originally intended to be used, be it inside Apple or otherwise, often causing serious design-level bugs, often resulting in major rollbacks. The lack of any QA work is also very apparent, especially in recent years, where simple bugs and glitches are allowed to go through, let alone major rewrites. The only tool to counter this is radar/feedback assistant, which are tools well documented in their inefficiency from current and ex Apple developers themselves. I don't think this should solely fall on Apple developers, of course. As is often the case, such inherent problems in culture are due to lackluster management. This has been hinted at by many ex Apple engineers.
- chmaynard 7y agoAgreed, but at least Apple managers have great hair.
- ohithereyou 7y ago>Apple developers rewrite software in a butterfingered way on a consistent basis, creating software that lacks any design, missing large chunks of functionality and often lacks any kind of polish or coherency. The day they completely gimped Disk Utility is the day I started looking to move off macOS.
- LeoNatan25 7y agoYes, that's a good example. A software that was messed up, had crazy security bug and is still lacking. Why was a rewrite necessary? How cynical would be to say "the engineer of the day fancied Swift more than going into old code"? Sadly these days, I'm not sure this is so far-fetched.
- dreamcompiler 7y agoThis explains something I ran into recently. I stick with older versions of macOS for a variety of reasons, and El Capitan cannot rename or delete a file that has a NUL character in its name. Earlier versions of macOS could do so. This implies they rewrote the unlink function and made a rookie C programmer mistake in the process. Why rewrite unlink? The mind boggles.
- okdana 7y agoWhat do you mean 'rewrite unlink'? As far as i know, none of the standard POSIX/C file-manipulation functions, including unlink(2), have ever supported NUL bytes embedded in file names. Path arguments to those functions are all NUL-terminated strings, as are exec*() process arguments, so it would be impossible to manipulate or reference those files using any of the standard APIs or command-line tools. I can't think of how macOS could have previously supported NUL bytes in file names without those files existing in a separate universe that can only be interacted with by software that uses some proprietary Apple API. And if that's really how it worked, i feel like that was the real mistake all along...?
- klodolph 7y agoWell… I do not know details here. But there is, in fact, a separate “proprietary” API for interacting with files on macOS, provided for backwards compatibility. This API exists because the POSIX file functions do not provide similar enough functionality to the earlier API. This API is called the File Manager. Documentation is hard to find on the Apple website, but you can find some fairly extensive documentation for it in the old Inside Macintosh series (and sometimes you can dig HTML or PDF versions of this on the Apple website, but whether this is possible, and the exact location, seems to change every couple years). The way you would specify a file on old versions of macOS used a length-delimited string for the filename, and referred to directories by 32-bit IDs. This was packed into an FSSpec structure which was passed to most functions which operated on files. Being length-delimited, it is possible to pass a NUL byte. From what I understand, the macOS kernel exposed some additional entry points for these functions to operate correctly.
- nornagon 7y agoFWIW, I created a mirror of all of Apple's OSS releases on GitHub here: https://github.com/apple-opensource-mirror https://github.com/apple-opensource-mirror