Comment by Endy
7 years ago
"No longer". That's a very bad sign about QC and how confident the major retailers are that we're not going to switch to a different and more stable OS. The problem is that they're right, because Unix/Linux/etc was never and is not meant to be a single-user desktop home OS for the general public. Of course, random broken updates completely bricking your system should be no surprise to those users either.
Anybody got a suggestion for an OS I can use that exists on hard media, doesn't use kernel or base OS code that's been distributed digitally and has optional completely non-destructive updates via hard media no more than once a year, so that I'm not feeling like I'm trying to hit a moving target with the stuff I want to use, and it either works or not until a year later?
It certainly brings the suitability of Windows for production use into question. Not only this bug, which is outrageous, but the direction of Windows 10 in general before this.
Try OpenBSD.
I actually used it as a desktop OS in the early 00s, it was a very relaxing experience. It was years behind FreeBSD (which was years behind Linux, which was years behind Windows...) in terms of device drivers, but it was incredibly stable, minimal, beautifully documented and well-organized OS that was a joy to work with.
As an example of what "stable" means here: OpenBSD has a port collection, but it's an implementation detail that you're not supposed to use. Instead, when a release is created, all ports are built and tested to guarantee that they work. Then, they're not updated. Ever. You're supposed to upgrade to the next release yourself, and if you don't, all the packages available to you right now will be available and working the same way 20 years from now.
(Of course, there is a -current version you can use to get more liberal update policy.)
I don't remember all the details, but it sounds like it would fit your requirements very well.
I may have to try it then. How does OpenBSD name drives? Logically, like A:\ & B:\ for floppies or removable media, C:\ for main HDD, D:\ for main optical media or second HDD, E:\ for second optical media, etc? Or does it name in the incomprehensible method Unix & Linux use that have no application to what I'm trying to do with my life?
Physical devices are named based on the driver used. Floppy is fd0 (fd1, fd2, etc. if you have more than one), cd drive is cd0 (cd1, etc.). These are generic drivers, with hard disks it gets more specific - for example, a Western Digital disk would be called wd0.
These physical devices are "mounted" into a single logical filesystem hierarchy, starting at /, with user data in /home/<username>/ (vs. C:\Users\<username>\). Various disks (and partitions) can be mounted basically anywhere in the filesystem, under any name you choose. You can mount your floppy as /A, your CD as /D and your other disk as /G and the OS won't complain at all. Traditional location for that is /mnt (ie. /mnt/G), but it's just a convention you don't have to follow.
You can learn about all the other places in the filesystem, basically what goes where, in `man hier`[1] It's like reading about the internal organization of C:\Windows, though, and can be ignored for the most part: normally, you stick to your home directory for everything and can organize it any way you like. Some desktop environments will create (and display in a special way) a more traditional set of folders, Documents, Downloads, Music and so on, but they will all be inside your home directory. You can mount your other drive (or a part of it) as your Music folder if you want, too.
As a side note, I'm not sure I would call Windows naming convention logical. It's just what you're used to. A single filesystem for everything vs. a separate filesystem for every disk/media is really a minor difference, just like the different path separator (\ vs. /) is irritating for the first week and then you stop thinking about it.
A warning, though. OpenBSD is not a desktop-oriented or newbie friendly OS. It's meant for servers and power users. There's very little hand-holding - even the installer has no GUI at all, not even console-based, it's just a command line where you type responses to printed questions. The primary interface to everything is the command line. The OS is focused on security and stability, not convenience. That being said, the system is incredibly well-documented and discoverable, and the community is very nice and welcoming, so it's not hopeless; just be aware that you're going to struggle for quite a number of hours before you make it work exactly the way you want it. The upside is that when you finally make it, it will stay that way. Forever.
[1] https://man.openbsd.org/hier
A live DVD of some Linux variant (Puppy or Alpine come to mind)? I'm not sure that I would want to run it for that long unless you figure out some way to run your browser off a different device, though; security updates aren't something that's safe to put off for a year.
Can download the binary Moz-distributed FF and unpack it wherever, then run $wherever/firefox. I think currently needs PulseAudio if you want audio, so that might limit what distro one can use. There can also be some traps with missing libraries, so very stripped down distro might be unusable. Otherwise, just check Moz website for updates every couple months?
At worst Linux updates might screw up your system and render it non-bootable. I'm not aware of any bugs in the past 20ish years where your home folder was deleted.
https://www.phoronix.com/scan.php?page=news_item&px=MTIxNDQ http://tldp.org/LDP/lame/LAME/linux-admin-made-easy/crash-re...
Apparently the bug doesn't actually delete your files that often, if you read the article. Also that second one isn't really something unique to ext4, is it? Seems like a universal problem.
1 reply →
Well.... there are a few really nasty issues that come to mind, usually surrounding filesystems. Not too long ago, if you were using a certain version of systemd on certain laptops and did a rm -rf / (which as a new user is not hard to mistakenly do), not only would you lose your files, your hardware would become unusable as well.
(Systemd mounted an EFI partition read/write, and nuking / also nuked critical firmware information)
Some cursory googling reveals some other issues, but everything I can find is either very old (ex. https://lkml.org/lkml/2012/10/23/690) or reference utilities that an average user won't have (ex. https://www.spinics.net/lists/linux-bcache/msg05290.html)
That said, given that Microsoft charges money for this, goes out of their way to render themselves not liable when their code breaks your shit, and doesn't test a use case that's not exactly uncommon, it's still unforgivable and not comparable to Linux.
Linux usually breaks when the hardware lies about having written something to disk or flushed the disk (both are more common on consumer hardware than you think, especially cheap SSDs), though I don't think any FS would fare well if the hardware starts lying to it.
There is also this[0] short blog post on FS reliability on Linux where they analyze the source code for cases where error codes are dropped (code like "if(err) { /* ignore */ }" counting as handled error here) and there was some significant amount of problems across all FS in the kernel.
Though, I don't think the other OS' out there are much better, FS are hard. Getting them right is harder. Getting them right while being compatible with existing implementations is very harder.
0: https://danluu.com/filesystem-errors/
2 replies →
Oh god, the EFI bug. I thought it was utterly absurd that some people on that mailing list defended that behavior and it really put me off Linux culture and the Linux community.
4 replies →
Linux also destroy intel e1000 nics for a while. And up until ext3 it lied about synchronous writes. Linux has been far from safe for storing your data.
1 reply →
If you are not experienced, you should not use root, because it's dangerous. Your experience confirms that. When you need to use root, plan for disaster.
You opened command terminal. You entered user mode using sudo. You entered command manually instead of using a user-friendly file manager, e.g. Midnight Commander. You disabled user-friendly interactive mode for rm. Literally, you said "system, delete everything and don't ask any questions". Yeah, sometimes sheet happens. I did that too once on production server with about a hundred of web-servers because I hit enter in the middle of the command. But I never blame system for my errors. Now I write tested scripts to make changes, use RPM to deliver updates, and use file manager to manage files manually. Learn from your mistakes. It's price for performance. For example, I needed to erase my partition recently. Using Linux, I erased it in less that minute.
5 replies →