KDE 3.3 UI, Evaluated By 7 Real Users

sgtrock writes "UserInstinct has an article that documents the reactions to KDE 3.3 by 7 users with limited or no Linux background. By and large, they found KDE intrguing, but far too busy. They all complained about some pretty basic UI issues; no stars echoed while the password was typed, anti-aliased fonts off by default, far too many options shown by default, etc. Most had minor UI issues as well; some of which were KDE specific, others that weren't. All in all, I would have to regard this test KDE 3.3's UI as somewhat disappointing. Especially since KDE strives so hard to provide a consistent and easy to use UI. IMO the single biggest issue is probably the overwhelming number of options that are exposed to a first-time user. Personally, I'm a guy who likes to tweak almost everything on a UI, so I love the way that KDE works. However, I have seen myself just how quickly beginners get turned off by the very busy option screens. There must be a better way of showing a sane number of options while still allowing power users to access the more esoteric ones. Anyone have any thoughts?"

Comments

Two tiered Configuration

[ADRA]

It's called Advanced... or else you could really break them up and call them Registry or gconf-editor

No standardized UI

[alatesystems]

The major problem that I have with using GNU/Linux as a desktop is the lack of a standardized UI. In Windows or Mac, no matter what program I'm in, I know based on UI guidelines and traditions that I am going to be able to use alt+f4 to close a program or window+r to run something, or command+f to open sherlock.

It is also a strength, because of the different choices developers have, but as a user, it is annoying to not have a consistant UI.

If you just use a mouse, it's not a problem, but the keyboard shortcuts are a big deal to me.

A lot of good programs are GTK and a lot of good ones are QT. You almost have to CHOOSE if you want to have a consistant UI.

Chris

Don't know where

[Apreche]

I don't know why people view KDE as the easy to use linux desktop environment. Xfce has taken that spot on my list since I discovered Xfce4 over a year ago. I've always viewed KDE as the complete and fully featured desktop environment that has a fancy working gui for absolutely everything and a matching integrated(looking) application for all the major tasks.

This would have been much more interesting if they had tested the users on 4 or 5 different linux desktops. Maybe Gnome, KDE, Xfce4, then throw in ion or ratpoison for fun.

I'm still waiting for some sort of objective comparison of major X11 environments against OSX and Windows XP. But I don't think that sort of thing is possible in this day and age.

Personally, while I like and have found good things in just about every desktop ( and I've used them all ) I've pretty much fallen for fvwm2. Fvwm2 allows me to customize everything by forcing me to customize everything. Hell, I had to configure the focus policies by hand. Of course, now that I've done it and made several backups of the configuration so I never have to do it again, I have created what is the closest I have ever gotten to my ideal graphical environment. It also looks really cool too.

Re:Don't know where

[moreati]

Correct, with sponsers such as HP, Sun & Red Hat, Gnome is aimed squarely at the corporate desktop where consistency and manageability are watch words. It seems also, that Gnome appeals to those who prefer simplicity (or usability, if you prefer) to versatility.

KDE is more volunteer driven, hence it aims to appeal to fellow developers and home power users, for whom configurability and features are wanted.

As a biased KDE user, I think it has the potential to be better than Gnome on the corporate desktop. However this study is right, there are too many 'in your face' options. I agree that KDE needs to cut back on the number of buttons, menu entrys, context menu entries and configuration options shown at the top level. Examples:

• The default konqueror toolbar has cut, copy and paste buttons - these are unnecessary and clutter the interface.
  
• The settings menu on most KDE menu is a mess, 4 different 'Configure &ltx&gt' entries in the case of konqueror, and what is 'Full Screen Mode' oding there? There may be a case for scrapping the menu entirely and placing the items in other menus.
  
• File context menu in konqueror is too overloaded, there are so many that it slows down 'quick access' to the commonly wanted functions. For instance there are entries for both 'Move to Recycle Bin' and 'Delete'. Only the first should be displayed by default, the latter perhaps appearing on 'Shift+Right Click', ala MS Windows.
  
• The number of KDE Control Panel panels needs rationalising, particularly sprawled, is the Look'n'Feel branch.

There are items in the study I think are dumb, KMail is a fine name for the KDE email client, much better than Kamel (or whatever the suggestion was).

The KDE project is making movements in the right direction, there is a nascent KDE-Usability project. Gnome has gone too far in usability through streamlining, KDE hasn't gone far enough, yet.

Of course all that's opinion, but I say it anyway.

Alex

Too many options

[xutopia]

"far too many options shown by default".

I don't want to make 101 decisions when I work on my computer. I want sensible defaults and not have to care about plethoras of unneeded options.

The best option?

[Acy James Stapp]

Don't have options. Pick something and stick with it. If it doesn't work, change it.

Real people don't like options. Weirdos like options.

Why one-size-fits-all?

[jeif1k]

There must be a better way of showing a sane number of options while still allowing power users to access the more esoteric ones.

Why does everybody try to be one-size-fits-all? Maybe infinite customizability is possible, allowing in a single code-base to be used by both novices and experts, but that may be a bad engineering tradeoff.

There is room on Linux for different desktop environments for novices, experts, Windows refugees, Macintosh refugees, Motif refugees, keyboard fanatics, disabled users, kids, gamers, computer labs, and many others.

I think XFCE and Gnome are a good example of two desktop environments that share a lot of code but are targeted at different user communities.

I use both Gnome and KDE. Gnome strikes me as a more "UNIXy" kind of desktop environment, but KDE's strength could be that it feels more comfortable to people coming from Windows or Macintosh. Maybe that's the user community KDE should target more aggressively, rather than trying to be all things to all people. That would mean removing some customizability from KDE, making its interface more Windows/Mac-like, and simplifying the UI.

Simplistic 80/20 advice...

[eyeball]

Take the most common 20% of the options, the ones most likely to be used by the 80% of the users, and present those. At the bottom, have an "advanced" button that reveals the other 80% of the configuration options.

an 'advanced options' checkbox?

[vjl]

Why not make some of the settings in KDE have an advanced options checkbox, like Video Lan's VLC media player has? When you open up the prefs in VLC, there are several settings, but some are hidden unless you click in the checkbox labeled, "Advanced", and then you get to change more options.

It seems like this is a better method than eliminating features from a preference dialog box [ala Gnome].

/vjl/

HID 101

[bay43270]

The term is Progressive Disclosure. It's basic human interface design. This is why open source projects don't pass these studies with flying colors - rather than ask UI experts to solve UI issues, we ask slashdot. Then we all hash it around until one underrated post manages to reinvent the wheel.

Here's a better question.. how dow we sell open source to HID people?

Re:HID 101

[Brandybuck]

First of all, progressive disclosure already exists on the KDE (and all other) desktops. That's why it doesn't have everything in one single menu, or why the interface isn't a single command entry widget. Instead we have multiple menus, context menus, sub menus, cascading menus. We have dialogs and toolbars. We have tabbed dialogs and advanced dialogs. People who complain about this aren't saying it doesn't exist, but merely that it hasn't been implemnted to the degree that they want. We can always politely discuss to what degree you want a progressive disclosure, but if you insist that it doesn't already exist, then we have nothing to talk about.

Second, please show me the progressive disclosure on my automobile's interface. As near as I can tell, every driving control is initially displayed, and remains displayed for the duration of my driving. Perhaps the concept of progressive disclosure isn't a basid human interface design. Hmmm, it doesn't exist on my microwave oven either.

Are "Advanced Options" really advanced?

[digitect]

Many times, options are offered as a lazy hack instead of making a UI really usable. If 9 out of 10 users need some advanced option checked, why isn't it on by default? Why is the alternate behavior even needed?

I also think that checkboxes are sometimes offered by the programmer who can't decide how he wants his app to behave. By offering multiple behaviors, he escapes having to commit to one or the other under the guise of offering the user more. But unless you have nothing better to do than to twiddle checkboxes on and off all day, most users won't touch the defaults, meaning that all those options the programmer thought he was offering are in fact unused.

The better way to UI design is to decide what the task is and then to offer a default behavior that best offers it. If there is indeed an alternate that could be prefered by some, perhaps multiple options can be ganged together and toggled with one switch. I do this in my text editor configuration: The option to type with the words wrapped at the window margin is coupled with not setting an auto-textwidth (auto-truncation of lines at a given length) and with using tabs as true tab characters. The opposite is to auto-truncate at a set width, ignore the window margin, and use spaces instead of tabs since the two most basic behaviors are to enter text to preserve line formatting and to ignore it.

Most options aren't really options. And you have to realize that those offered have the potential to confuse more than to help. Once you develop a highly conservative perspective about offering options, you begin to value each one offered more. Besides, if the so-called power users really are, let them fiddle with configuration files, registries and source code like they claim to understand. Just don't bog down the average user!

My take on a few things...

[7-Vodka]

1. no stars echoed while the password was typed:
   This is the desired behaviour from a security point of view. It's something that people can get used to fast too, it only confuses on the first try. And it's only the default, you can change it if you think security is stupid, it may even be slack-kde specific default but I wouldn't recommend changing it.
2. anti-aliased fonts off by default:
   Yes they should be on by default
3. far too many options shown by default:
   This is where it gets interesting. Do you want to please people who will only ever use kde once, or do you want to please people who will use kde all the time?
   Yes, kde has enough options to be confusing on *the first time you use it*. This is the only sample group they tested. Why don't they have these same people use kde everyday and come back and say if they still want all the options hidden.
   This is where the gnome ui designers went wrong. They took away all the options and messed up their desktop trying to be the best DE for first day users. Then the users use it for a few weeks and are no longer first time users and switch to something more meaty. This is a complicated thing to think about and I don't think it's worth it to change the interfaces based on sloppy investigations. You have to look at much larger groups of users and see how you can please them all. I'm thinking it's impossible to please everyone, but it's possible to please 90% of new users and 90% of experienced users as well.

Besides all of this, isn't there something called kde-kiosk that lets you turn off all the options? isn't this something done at a distribution level rather than kde-project level?
I say let the kde group fix upstream fixes and let the distributions fix their own bugs. IE if you're a distro targetting users who've never used kde, you fix the ui the way you need it, while if you target advanced users, you fix the ui the way you need it. Let the kde project make kde more maleable and let the distros do the final moulding of the user experience.

Re:My take on a few things...

[BenjyD]

far too many options shown by default:

An example of KDE having too many options:
There is actually an option "draw borders around maximized windows" and it's on by default. It's not actually called that of course, it's called "allow moving and resizing of maximized windows", but it has no effect on moving any more. Why would anyone want a border around a maximized window, making the fitt's law violations even worse? Let alone an option for it. Why not just replace the option with a sensible default.

Mostly distribution issues

[brunes69]

no stars echoed while the password was typed,

This is a global option in KDE. It's up to the distro how they want their defaults. Personally I like the stars *off* so a casual passer by can't see how long or short my password is.

anti-aliased fonts off by default

Again, a distribution thing. If you compile KDE from source yourself you'll find them (surprise!) on by default.

Kopete send messages by pressing enter

Again, default config thing a packager could easily change. And in current KDE CVS it is already changed.

far too many options shown by default, etc.

A distro could easily change the default KControl link to point to settings:// instead, which is far less confusing.

Mostly an uninformed article IMO. If the reviwers want to test the newbie usability of *KDE* itself, not of whatever distro they happen to be on, they should at least spend some time learning how to do things in KDE so that they can set up the system to be newbie-friendly *before* the newbies test it.

Bad study

[Brandybuck]

I have to say this yet again, because I think it's terribly important for people to understand: studies like this are merely trivial.

While it is a worthwhile goal to create a desktop that creates an initial good impression on a newbie, it should not be the primary goal, or even a major goal. That is because the newbie will NEVER be the primary user! As heretical as it sounds, no one ever stays a newbie forever.

Usability studies need to be directed towards intermediate users. Those who have been exposed to the system for more than thirty minutes. I don't give a stack of meadow muffins how easy a system is for a newbie if it turns out to be difficult for me. I don't want to make a choice between the newbie and the expert, but if a choice had to be made, I would rather reward my long time loyal customer than someone just sniffing around for a good deal.

So here's how to do the next study. Don't base it on the anecdotal statements of a few people. Give the test subjects the destkop a week to get acclimated with the destkop. Don't ask them their feelings but observe their actual behavior (use a video camera if you have to). Try to get a control group or control desktop in place. And don't confuse the user with a usability expert

Re:Bad study

[Minna Kirai]

As heretical as it sounds, no one ever stays a newbie forever.

But everyone was a newbie once. You can't ignore either side of it: design only for newbies, and only newbies will use it. But design only for experts, and NOBODY will ever use it.

Every UI designer must aim for the right balance of accessibility and power.

I would rather reward my long time loyal customer than someone just sniffing around for a good deal.

That makes sense if you actually have those long-term customers. Microsoft does; KDE does not.

Non-monopolist UIs have a greater need to focus on new users. They're a bigger part of KDE's potential audience. Most computer operators have used MS Windows, but only a slender fraction have any KDE experience.

PS. My primary workstation runs KDE exclusively. Even though I hate single-clicking to launch icons.

Re:MERCY

[be-fan]

Please, MERCY, stop putting pressure on the KDE developers so that they turn KDE into a Windoze kind of sh*tty UI, just in order to please the newbs.
The problem now is that KDE is too much like Windows. I've been a KDE user since 1.x (and am typing this in KDE 3.3.0), and even I can admit that KDE suffers from largely the same UI problems that plague many Microsoft apps.

Specifically:

1) Lots of toolbars full of lots of cryptic, hard-to-decipher icons. It was the UI-disaster called MS Office that made this style of app popular. Most KDE apps copy this style. Konqueror is laid out like Internet Explorer (> 10 icons in toolbar), instead of like Firefox or Safari (~5 icons in the toolbar).

2) Ultra-long menus and context menus. It's Microsoft and Windows that popularized the 20-item context menu, which KDE faithfully copies.

3) Multiple places to configure functionality of application. KDE's settings menu copies this nicely, though it's not as bad as what Microsoft does with Visual Studio (which puts configuration both under the File menu and the Tools menu).

My thoughts

[rRaminrodt]

I read this article before it hit /. so I'm a little surprised by how many people are posting the same comment (to add an advanced dialog mode) at the same time I agree to some degree. If you've used KDE lately you know there are quite a few dialogs where there are "advanced" buttons. For instance, the "Style" gui window has only 4 widgets, not counting apply/cancel. One of them is an "advanced settings" button.

Of course, like anything there is a lot of room for improvement. I think as more apps are moved over to the new KconfigXT system it will be easier for folks to tweak and improve dialogs. This library lets people describe the settings instead of coding them into a dialog.

I also don't think the invisible password thing is a KDE default. It might be slakware's. I compiled (emerged) kde from source and my login screen shows the stars.

a computer is like a car

[LeninZhiv]

Studies like this are wrong-headed: car lovers would ridicule the suggestion that all models should be made alike, maybe we should think of our computer's WM 'dashboard' the same way.

Simplified, standard business UIs are like station wagons--practical, easy to use, automatic transmision all the way.

KDE is more like a Porsche--fine tuned control, manual transmision, yet still fully polished, appealing to a broad category of enthusiasts who take their vehicle more "seriously". You can argue all you want about manual transmission not being necessary anymore, the people who drive these cars *want* it and don't care that it's less 'accessible' to others.

The smaller, more exotic WM's are like the cult classics--Ford Cobra, classic Mustang, etc. These are for people who have a fascination with mythical features that goes far beyond the desires of the general public. The fans of these cars look down on the mainstream "enthusiasts" as wanna-be's, and take pride in the long and esoteric learning curve that has allowed them to develop this closeness with their machine. Try explaining to one of these guys that an automatic transmission station wagon is a superior design!

My point is that none of these groups are 'right' or 'better'. The only way to go wrong is in trying to be all things to all people.

(With apologies to "In the beginning was the command line" for stealing a good metaphor.)

usability is no longer the issue...

[dash2]

... Well, it is still an issue for KDE, because they haven't made the (quite far-reaching) changes to their priorities needed to get developers thinking about usability. Gnome have done this much better IMHO, and KDE has even suffered because lots of the software geeks who want "options everywhere" have moved to KDE... making the user base even less typical of the Joe Average user.

But in the long term, I am sure KDE will make strides - the Akademy seems to have focused on this area, and KDE are a great project which can develop at lightning speed.

In any case, the real issues for the "free desktop" as a whole are now elsewhere. This would be my first approximation:

* office suite. OpenOffice is good... it's almost as good as MSOffice. But it needs to be BETTER, in order to get people and institutions to switch. After all, 90% of ordinary people's computer time is spent working with office-type applications.

* hardware. I have this sweet Gnome desktop... but I have just spent a day trying to get my printer to work, with NO help from the GUI. Hotplug and HAL are a start but it's not just about recognizing hardware, it's about an easy way to troubleshoot it. Is my problem hardware related, or a kernel problem, or a driver, or is it CUPS or foomatic? I don't know!

* Help systems. Gnome and KDE have great manuals for individual programs. They don't have task-based HOWTO guides, or troubleshooting guides, saying "how do I connect to my friend's workstation" or "I can't get online, what's wrong?"

* Stack integration. We have a great kernel. And we have (at least) two great desktops. And in the middle we have... well, an assortment of GNU and BSD tools, distro-specific stuff, little programs that someone wrote, each with their own config file.... Again, the Novell stuff and the freedesktop stuff isn't a bad start, but we really want *one single way* to manage startup services, for example, so that GUI tools can be written which will work with every distribution (without writing 10 distro-specific backends). The advantage to this is that when there is "one true way", it is much easier to build stuff that can assume the "one true way" is being used. (Note: this is perfectly compatible with diversity, experimentation, Gobolinux etc. etc. - just remember, experimentation can be aimed at *improving the next generation of the standard*.)

* In general: "finish". This is related to the stack integration point but it is more abstract. Let me give you an example. I use Debian. I need to change the time, due to an obscure 2.6 bug relating to APM which makes my clock run at 4x speed. I read "man hwclock". It tells me about how to use hwclock to set the time, and that I can automatically adjust to make up for a slow clock. So I do that. Then, by chance, I read "/usr/share/doc/util-linux/README.Debian.hwclock.g z"...
and it tells me all about how hwclock is dangerous and must not be used! You see? The same distro is giving me inconsistent information in two different sources!

Another example: network interfaces. I can access these in so many different ways. "ifconfig". "ifup" and "ifdown". The scripts in "/etc/init.d". And each of these different levels is built upon the others... and I often don't know which one to go in at... and if I use one level, I may break the assumptions of the other level. Similarly, how do I configure samba?

So IMHO "finish" is the big one. It's about turning a big heap of user-contributed software, into a single consistent centralized system. That's where distros come in, but at the moment, those distros aren't really doing their job. They are just whacking out 6-monthly releases with the latest cool stuff all chucked in there. Debian is better - kudos for a slow, careful release cycle and a high standard for package maintenance - but still, as a desktop, it ain't there. We should be aiming for a distro release to be usable for *years*, not months - my Mum still uses Windows 98, why is this seen as "old-fashioned" if it