Why Linux sucks

[Harovan]

SuSE, RedHat & Co. annoyed the hell out of me for several years. They're bloated, unstable, expensive (yes! having to upgrade thrice a year to get 1000 bugs fixed and have 2000 new created is expensive) and their installation and admin tools suck. So I switched to Debian stable a couple of years ago, and I would never use another distribution again, when it comes to Linux. I'm not much of a version junkie anyway, and I use Linux only as server. BSD fans will say, that xxxxBSD is yet better on the server and I agree with them, but the time I set my server up I had no BSD at hand, found only an old "potato" binary-1 CD, it worked fine and I just don't want to change a running system now.

Linux will continue to be inferior to Windows at the desktop as long as its GUIs are X based. Unfortunately, there is no similar subsystem in sight. My own experiences with both systems as an application programmer say, that Linux has an edge when it comes to networking abilities and performance, while Windows is far superior in process and thread handling and of course in its GUI part. May be Linux will gain terrain, when it finally scraps these ugly clone() based POSIX threads, but when you have portability issues and are stuck with them, you're just screwed.

[Urban Ranger]

Originally posted by Sir Ralph
Linux will continue to be inferior to Windows at the desktop as long as its GUIs are X based. Unfortunately, there is no similar subsystem in sight.

Why's that the case? X-Windows allows you quite a bit of freedom in designing the windows manager. In fact, decoupling the GUI from the OS is a strength, not a weakness.

Originally posted by Sir Ralph
My own experiences with both systems as an application programmer say, that Linux has an edge when it comes to networking abilities and performance, while Windows is far superior in process and thread handling and of course in its GUI part.

Are you sure? My previous experience is that task switching under Windows is a lot heavier than *nix.

[Harovan]

Originally posted by Urban Ranger
Why's that the case? X-Windows allows you quite a bit of freedom in designing the windows manager. In fact, decoupling the GUI from the OS is a strength, not a weakness.

If X were stable, I would agree with you. Unfortunately, it is not. It may have changed, I used X only occasionally in the past 2 years, but somehow I doubt.

Are you sure? My previous experience is that task switching under Windows is a lot heavier than *nix.

Spawn 50-60 threads from the same application on the same machine under Windows NT/XP and Linux, and you'll see . Linux was practically inoperable after it, Windows only a bit slower. I've done some extensive testing on this. I've been told to use another thread subsystem, but I'm stuck with POSIX for portability reasons.

[Urban Ranger]

Originally posted by Sir Ralph
If X were stable, I would agree with you. Unfortunately, it is not. It may have changed, I used X only occasionally in the past 2 years, but somehow I doubt.

Explorer (Windows' GUI) isn't particularly stable either. Even though Windows keeps restarting Explorer whenever it crashes, so the system could be used instead of staring you with a blank screen, this whole design makes remote access very resource intensive and slow.

[Harovan]

Originally posted by Urban Ranger
Explorer (Windows' GUI) isn't particularly stable either. Even though Windows keeps restarting Explorer whenever it crashes, so the system could be used instead of staring you with a blank screen, this whole design makes remote access very resource intensive and slow.

Agreed. Remote access falls under networking abilities, where NuxNix IMHO has an edge anyway.

[Angelo Scotto]

Originally posted by Ari Rahikkala
Weell... it is Unix-like, but the relationship between WXP and WNT is quite different from the relationship between Linux and the original Unix. Linux was written from scratch to be compatible with Unix; XP is a point release

Well, the fact that Linux was written from scratch has nothing to do with the fact that it was (and is) obsolete as a Kernel (i totally agree with Tanenbaum on this point). Linux is a monolithic kernel (an hybrid one since today it has LKM) based on Unix monolithic kernel and those type of kernels are considered obsolete since late '80, NT Kernel instead is a micro-kernel (like HURD will be), even AmigaOS was technically more advanced than Linux and it was up and running in '85! (Why do you think that Amiga community rejected the idea of using Linux as kernel for future Amiga OSes and decided to keep developement of AROS? and why GNU continued the developement of HURD even if Linux is available?)
Even Torvalds planned to release Linux as a quick hack to run GNU programs while waiting for the HURD.

Originally posted by Ari Rahikkala
HURD might become a revolutionary advance instead of only an evolutionary one, but last I checked its performance and stability were abysmal and it supported, well, basically nothing. I'll have them call me back when they show that HURD is ready for real use.

Well, developing a micro-kernel based on Mach is without doubts a huge effort, and without a company behind you it's almost an impossible task (and it makes monolithic kernel developement a trivial task)

Originally posted by Ari Rahikkala
As I said in the first post, I don't really care about the "confusion between the Linux kernel and the inexistent Linux operating system". You know, I would actually prefer to just call the operating system GNU; Unfortunately nobody would know what I'm talking about... "GNU/Linux" is too long to use daily as a name of an operating system, especially if you're supposed to pronounce the slash aloud... so, anyway, the thread title was referring to GNU/Linux distros really.

I wasn't doing a debate about names, i wanted to make clear that when i say "Yes, Linux sucks", i'm referring to the kernel, and not to the whole system.
You say that the whole system sucks, well, it depends what you expected from it, the best thing of GNU/Linux(Linux,GNU call it as you want ) is the fact that you can have a free operating system with full sources, completed by free compiler, drivers, applications, etc... obvioulsy if you expect it to be a panacea you'll be disappointed (i know that Linux evangelists promised a lot, but you shouldn't give them more credits than MS Evangelists or Apple Evangelists and so on...).

Originally posted by Urban Ranger

It is not unusual for an application to crash NT.

I repeat, it IS unusual thanks to memory protection mechanism of NT Kernel.

Originally posted by Urban Ranger

but the big thing are drivers.
For some unfathomable reason, MS decided to stick drivers into the NT kernel. Needless to say, that caused a lot of headaches

I'm afraid i don't understand your point:
The unfathomable reason is stability and security: have you idea of what mess will cause allowing application code to communicate directly with the device?
Also in Linux (and i think every OS in the universe) device drivers works in kernel mode, when you load a LKM at runtime, the code is "extending the kernel" and became part of the kernel so, if it is buggy, the whole system will go down; windows does the same with Vxd (on windows 9x) and with "Kernel Mode Driver" (on windows NT) so, not a big difference here.

[Ari Rahikkala]

Originally posted by Sir Ralph
Spawn 50-60 threads from the same application on the same machine under Windows NT/XP and Linux, and you'll see . Linux was practically inoperable after it, Windows only a bit slower. I've done some extensive testing on this. I've been told to use another thread subsystem, but I'm stuck with POSIX for portability reasons.

Yep, LinuxThreads was a hack. Sucked both performance-wise and behaviour-wise. Newer glibcs and kernels are better, though. Have a look at http://kerneltrap.org/node.php?id=422 , for example.

As for X... yes, it sucks. But you know what? The way how apps use X sucks even more. Ever seen how Qt apps are rendered on X11? I have. Qt basically renders everything as pixmaps itself and sends them to the server. Sure, it means that there's not much stopping you from running KDE on a dumb framebuffer, but it also means that running KDE on top of a featureful graphics server like XFree is one goddamn shameful waste of resources. Gtk+ is a bit better, AFAIK, but still not great. On the other hand, it's no wonder everything needs to be done client-side - XFree86 development is so slow and unresponsive that depending on features being made supported on the server is not a way to stay alive for a long time.

You say that the whole system sucks, well, it depends what you expected from it, the best thing of GNU/Linux(Linux,GNU call it as you want ) is the fact that you can have a free operating system with full sources, completed by free compiler, drivers, applications, etc... obvioulsy if you expect it to be a panacea you'll be disappointed (i know that Linux evangelists promised a lot, but you shouldn't give them more credits than MS Evangelists or Apple Evangelists and so on...).

Weell... that's a point. But I never claimed that other OSes didn't suck .

[Harovan]

Originally posted by Ari Rahikkala
Yep, LinuxThreads was a hack. Sucked both performance-wise and behaviour-wise. Newer glibcs and kernels are better, though. Have a look at http://kerneltrap.org/node.php?id=422 , for example.

Sounds very promising, 100,000 threads started and stopped in 2 seconds. The mentioned 60 stock POSIX threads were started in about 7-10 seconds, and after this it was hard to gain the attention of the machine, it was busy with practically nothing (except printing characters on the screen and then sleeping). I had heard about these threads, but I can't use them yet. Perhaps after they make their way into Debian stable, but this may take 2 years.

[Agathon]

Originally posted by Asher

Now, as you can clearly see, any argument stating that the top of the screen is better is not an instance of this law at all, but rather an interpretation, one that, when taken into account of the big picture, is rather incorrect IMO.

Edit: BTW, the argument for the single-menu-bar-at-the-top comes from the idea that the top boundary becomes infinitely large, which means the access time is theoretically far less (maybe this is where he pulled his 5x figure from).

I know that's where his argument comes from - I can read and you are still wrong. But it still doesn't alter the fact that you clumsily relied on a site which takes it's version of Fitts' Law from a site you trashed the other day.

You contradicted yourself, plain and simple.

1) Studies have shown people do not overshoot much at all, particularly once they've used computers for a year or two

But that's not his point. You simply go slower in the latter case. This is not technical, nor does it require a great degree of mathematical knowledge, it's just common sense. Anyone with half a brain can see I'm right on this one - after all they can test it for themselves.

2) The single-menu-bar on top drastically increases the distances required (and clicks required) when multitasking, which throttles up MT in Fitts' law

And I pointed out what was wrong with this argument in a previous post. The fact is that most of the time people use a menu within one application. Switching between menus is much less frequent. Hence any time saved by having two menubars on screen is less than the time saved by having one menubar at the top. Add to that the wasted screen real estate and you have a clunky, useless interface.

[Agathon]

Originally posted by Urban Ranger

This, of course, runs smack into a linked site, which states:

Did you see how he tried to wriggle out of this one?

[blackice]

Here is an interesting article by Andrew Grygus, it is long but worth the read.

http://www.aaxnet.com/editor/edit029.html#altpath

Linux is coming into it's own and will get better with time, don't toss it out yet. All one has to do is look to the alternative to see it is a matter of nessesity that it does get better.

[Whaleboy]

I think what has become clear is that each OS has different "charms" as it were.

There seems to be something of a majority consensus that Linux is more stable, and faster on older hardware, more secure (which is in fact shown by all the bugfixes, not disproved by that). Windows is undoubtably easier to use, compatitble with more stuff, and works out of the box, whereas linux requires configuration to get the system working. Windows is also improving in stability, although its development model will prevent it from coming close to Linux.

The constant bugfixes are indicative of a good code-autiting system, and not keeping problems under wraps like a certain company based in Washington state. Thats one of the reasons I like OSS. You dont have to install the fixes if you dont want to, the fact that there are many and more frequently is not indicative of poorer code, merely a better system of discovering and fixing bugs.

Pretty much the same goes with BSD as Linux, although greater stability and speed at the expense of ease of use and "handholding".

Each system has its charms, and is best for certain people. It is somewhat silly of people to recommend that "all should switch to linux" or "all should switch to windows", because it works for them. Some people want different things, like I want a powerful, secure, fast and stable desktop for work and internet, and I dont play too many exotic games, so *nix is perfect for me. I dont care too much for ease of use, as I dont have a problem with being left close to the silicon.

I recognise that for others, windows is a better option. Neither side is holier than thou, and contrary to some of Ashers bull, Windows has many faults when put up against *nix. Few can deny the rise of linux, and the fact that MS's monopoly days are numberered, and I think that would be better for all in terms of competition. Microsoft will still exist and provide Windows (probably), and thats all the better for those that need MSware. I just hope that they dont try to impede others.

If people want to compete with other OS's, write better code and advocacy. Dont flame people with bull and pollute your karma, doesnt do anyone any good.

[Ari Rahikkala]

Just a couple of quick comments, Asher can do the flaming...

Originally posted by elijah
The constant bugfixes are indicative of a good code-autiting system, and not keeping problems under wraps like a certain company based in Washington state.

Actually... no. Since the point of auditing code is that audits should be done primarily before releases, less bugfixes means either better auditing or more laziness/secretiveness. See OpenBSD - a prime example of good code auditing.

Thats one of the reasons I like OSS. You dont have to install the fixes if you dont want to

Er... that's good, I guess, but compared to what?

[Asher]

Originally posted by Agathon
I know that's where his argument comes from - I can read and you are still wrong. But it still doesn't alter the fact that you clumsily relied on a site which takes it's version of Fitts' Law from a site you trashed the other day.

You contradicted yourself, plain and simple.

*sigh*
Apparently you can't read, or you're relying on other people not to read and you hope by continually repeating "you contradicted yourself" and "you're wrong" that you'll eventually be correct.

The fact is, you misunderstood what Fitts' law is, and your only argument stems from the guy who created the MacOS menubar and claims it's "5x" faster without any backup for that.

But that's not his point. You simply go slower in the latter case. This is not technical, nor does it require a great degree of mathematical knowledge, it's just common sense. Anyone with half a brain can see I'm right on this one - after all they can test it for themselves.

You're changing your argument on this one -- you used Fitts' law as a reason for this before, which is nonsense as I've tried to explain to you. It is true that people with inadequate mouse control will find it slightly faster, but the simple fact that the menu isn't as intuitive when it's not attached to the window it controls, the fact that the mouse cursor is usually in the middle of the screen and the menubars are not always in maximized windows, etc. -- they increase the time it takes to get to a menubar. Not to mention multitasking slowdowns with the 1-menu approach.

And I pointed out what was wrong with this argument in a previous post. The fact is that most of the time people use a menu within one application.

That's not countering my argument, that's finding excuses. Ones that aren't very valid, IMO.

[Asher]

Originally posted by Agathon


Did you see how he tried to wriggle out of this one?

Tried to wiggle out of that one?
Do you want me to post the Microsoft.com articles on UI and present them as absolute fact and watch you "wiggle out of them"?