German firm Relevantive has performed a formal usability test of Linux as a desktop OS (SuSE/KDE), with a control group using Windows XP. They found that user, who presumably had some familiarity with Windows, took slightly longer (almost equal) to perform an identical task in Linux and estimated that it would take a week for them to achieve familiarity with the new system. The study also identified several problems with Linux as a desktop system, mostly around the UI.
The best comparison I’ve seen so far. I am really astonished how narrow margin there was in between SuSE 8.2 and Windows XP. With this ease of acceptance already today, SuSE will surpass Windows in the coming years as a cheap, versatile, easy-to-use and stylish alternative.
SuSE might surpasss XP, but don’t forget that Longhorn is jsut around the corner in 2005.
Longhorn also doesn’t guarantee that Windows will be any more usable then it is now. It could be better, or it could be worse.
If you take a Linux distribution 5 years ago and compare it to Windows as it was 5 years ago, and then compare Linux distributions and Windows today, you see that the gap between them has almost gone away. Windows has improved a lot, but it has not been able to keep the pace. There is a good chance that by 2005 Linux distributions are so viable that Longhorn fails to attract audience in a way that it needs to.
Then the results might of been different
IMHO, one of the reasons why linux itself is successful is that it is it’s own design. It doesn’t try to be another OS. Sure it aims for POSIX compliance and it is very unix-like in appearance, but under the hood it is it’s own beast. On the desktop end of things, I am disconcerted that it seems like Linux is always playing catchup on the desktop front.
All of this is relatively moot, because there aren’t enough apps yet. I suppose it’s a catch22 – until there are apps, more people won’t use it(from the desktop-only realm); until there are more users, development companies won’t spend the considerable resources to port their codebase. Right now you either have to use winelib or rewrite an app using a cross-platform gui library or make a port based on another library. It’s quite an undertaking for a large app from most any vendor.
Drivers are a big mess. I just got Wireless 802.11g – no linux drivers. Graphics cards are always way behind in driver support or binary only crappy unstable releases… Again, not enough users to make the vendors support linux, and the vendors are still being stupid enough not to release full specs to the open source devel community to allow for the drivers to have a chance of being written. Obviously that’s not every company or product, but it means you have to shop carefully to know for sure something will work. That is just simply beyond a lot of the desktop consumer crowd who’s used to shoving any old piece of hardware into their machine and having it work. Windows has excellent driver support, insomuchas amount of supported gear.
Don’t get me wrong I think Linux may one day emerge as more of a contender for a mass market, but that will be only in a few years. For too many types of users, there are still important hurdles in way: drivers/hardware support, applications, ease-of-use/interface.
> Drivers are a big mess. I just got Wireless 802.11g – no linux drivers.
That’s not Linux’s fault however. It’s the hardware company’s fault for not providing Linux drivers.
> Then the results might of been different
Yes, it would have taken them a lot longer time to learn Gnome because it’s so much different than WindowsXP. KDE isn’t exactly like XP either, but I think this study shows that it’s certainly “close enough” to be useful for a large amount of Windows to Linux switchers.
Of course, it’s different for a completely new user or a MacOSX user, who might learn Gnome/RH faster than KDE/SuSE.
It’s an interesting report, it certainly shows how much progress KDE has made. But I feel it focuses too much on the desktop environment, IMO that’s a pretty small part of the system’s overall usability. Personally I like using KDE, it’s better than Windows in a lot of ways and it’s very easy to learn. But I’d estimate that 99% of my time is spent using apps, so obviously how well they work if far more important than the DE. Apart from a cursory look at word processing and e-mail this report barely looked at productivity software.
The main usability issues that stop me from switching to Linux weren’t even mentioned. For example, I consider the lack of consistent applications to be a much bigger problem than the DE issues this report highlighted. I found creating a simple newsletter to be very frustrating and a lot slower in Linux, simply because of the lack of cut, copy and paste of images between the apps I was using.
Obviously a lot of the usability issues depend greatly on what you want to do with the computer. The lack of decent dual headed display support is one of the main things that cripples my productivity in Linux. It’s something I’ve used for years in other OSes and I find it very hard to give up, even basic web browsing becomes a lot less enjoyable. The lack of hibernation damages my productivity too, it wastes a lot of time having to restart all my apps after rebooting. But obviously if you don’t use these features they aren’t a problem.
The point I’m trying to make is that this report is much too limited to really indicate how usable Linux is for any users who want more than basic internet use and word processing.
>> Drivers are a big mess. I just got Wireless 802.11g – no >>linux drivers.
>That’s not Linux’s fault however. It’s the hardware >company’s fault for not providing Linux drivers.
I don’t recall ever having said that it was “Linux’s fault” – I gave a clear reason why there aren’t drivers. The linux market hasn’t hit the critical mass to encourage enough hardware vendors to open their specs to OSS developers or write and release their own drivers. The fact remains that at the end of the day this presents another stumbling block to linux’s ability to entice more widespread use in the desktop sector, no matter who’s fault that may be.
To me this is a serious problem for the linux community going forward. The M$ establishment doesn’t really _want_ to have to support Linux, it seems. Look at how many companies’ products will become nearly useless in a Linux market: antivirus, software firewall, Norton Utilities type products… There’s resistance on many fronts.
There is one big-little thing holding back linux on the desktop IMHO and that’s xfree86.
There is one big-little thing holding back linux on the desktop IMHO and that’s xfree86.
I think XFree86 is quite a good implementation of X11.
A lot of extensions like randr are not available in other implementations.
> There is one big-little thing holding back linux on the desktop IMHO and that’s xfree86.
> I think XFree86 is quite a good implementation of X11.
> A lot of extensions like randr are not available in other implementations.
We’re not talking about the quality of Xfree86, it’s quite good. But the whole X architecture is outdated and could not keep track of modern requirements.
It was well suited what it was supposed for in the middle of the 80’s, namely network-transparent unified access to graphics workstations. It wasn’t simply designed for what we are using it today.
The quality of it doesn’t matter, it’s outdated. IMHO the main factor why we don’t have that many apps with a GUI is that there is no standardized way of creating one. The only way to write an app with a GUI that would work on every linux box with graphics (ok, i suppose every box with graphics running Linux has X installed) would be directly calling X functions. Now, you have the choice between two main toolkits, many others or writing your own (That’s exactly what Sun did with StarOffice and OpenOffice still has its own toolkit). That’s the reason why we don’t have that many programs with a GUI for Linux.
If you seriously want more GUI apps for Linux, take a look at X replacements like Fresco http://www.fresco.org/ .
Very good report. The article clearly says what the test was trying to accomplish and I feel they achieved it well.
The test was not about installing misc hardware it was about a managed environment such that might be found at work. Every pc is the same, same hardware, same software, Same OS version, same setup, everything. In this environment the pc’s will most likely be supplied by a Linux VAR of some type and all hardware and drivers and OS will most likely be already installed and configured.
In the case where a company is thinking about switching over to Linux they will have an established base of “older” hardware and most likely Linux will support these devices perfectly. In some cases better than Windows does. How many times have you had to throw old cards away because the “new” M$ Windows doesn’t support them anymore?
The only way to write an app with a GUI that would work on every linux box with graphics (ok, i suppose every box with graphics running Linux has X installed) would be directly calling X functions.
Wouldn’t it be better to create a lowest common denominator GUI interface so that all apps can write to it? Yes I know thats a pipedream,
Fresco is far from being usable !
Most of the time, it’s X that wait for the application and not the opposite. And don’t forget that when you drag a windows, millions of events are computed.
Well, in the test, I’m quite suprised of one thing. Linux is not quicker than Windoze … I’m suprised, because I’m running a machine with both XP and a Mandrake 9.1. And the performances of my Linux are around 20 % higher (according to Seti Working Unit calculation time; 6.45 hours under Linux , and almost 8.30 hours under XP). So, may be the applications are executing slower (and most of the one tested are running under JVM on Linux, but are compiled one on Windows…), but I affirm that the Linux Kernel is really more faster than Windoze Kernel.
More over, there is not only “Desktop” user on XP world. There is developper too. Just try to developp and compile under XP, and then do the same thing under Linux. You’ll see that Linux is leading (I think it is):)) (more power, more portability,more tools,more network integration and so on
)
and could not keep track of modern requirements.
Such as?
It wasn’t simply designed for what we are using it today.
And that would be? Aren’t we using it for networks transparent access to GUI applications?
Now, you have the choice between two main toolkits, many others or writing your own (That’s exactly what Sun did with StarOffice and OpenOffice still has its own toolkit).
I never understood why having more than one toolkit is a problem.
Applications can run side by side independend of the toolkit they use as all toolkits use X functions under the hood.
Just like Toolkits on windows use the same WinAPI internally.
That’s the reason why we don’t have that many programs with a GUI for Linux.
Why do you think there aren’t many programs with a GUI on Linux?
I know a lot of programs with X GUI and program without are usually ones that don’t need one (daemons, shell tools, etc)
Besides, I don’t understand why using toolkits would result in less applications. Noone would want to use base APIs, neither Xlib nor plain WinAPI.
IMHO using only base APIs would result in less applications, using frameworks does lead to more applications.
If you seriously want more GUI apps for Linux, take a look at X replacements like Fresco
So introducing another different system is a good idea but uing multiple toolkits on one system isn’t?
I am really confused now.
If it replaces X it will need some compatability layer for X applications.
If such a layer exits, how likely will someone develop a fresco application if she doesn’t need any fresco features and can reach other X11 platforms like unix by developing an X application.
how narrow margin there was in between SuSE 8.2 and Windows XP. With this ease of acceptance
Sure, SuSE Linux will surpass win XP ina year
Oh boy !
Actually the NT kernel itself is quite good, and one of the best around from a technological point of view. It has great concepts like the HAL, the ability to use different subsystems, it (apparently) allows to easily load a service in userland or kernel mode at your choice (microsoft choice that is), etc., all in all Dave Cutler did an excelent job.
The win32 subsystems on the other hand are pure crap.