Celebrating one month of KDE 3.3 out in the wild, userinstinct put together a usability review with user testing. "Based on feedback from our test group, the default settings for a number of KDE parameters differ from what is usually expected and desired by users. Providing better defaults would reduce the time users spend looking for configuration settings and would provide a better "out-of-the-box" experience."
Dot Categories:
Comments
This is absolutely right. I honestly believe that Classic MacOS is a shining example of this precise balance. The UI was designed to be functional, elegant, and clean. It wasn't totally minimalistic, but tended towards minimalism. For the majority of users, this was great. For power users, this was still great, because they had tons of extensions, plugins, and haxies to choose from.
Stars in password is a good thing. Could be disabled in KDE Control Center.
KDE Control Center is like TweakUI for Windows9x,
it is good for power users, but way too complex for simple task.
We need a "simpler version" along this tools,
for 'heavy common task' a strip down Control Panel would do it.
Advanced in that Control Panel could point to KDE Control Center.
You could also have an even more "complicated" version
that shows you ~/.kde/ config files ala "Registry Editor".
Therefore, if some newbie wants to do simple stuff,
he just go to his simple KControlPanel and fix it.
If some hacker want to freak is KDE, he already knows
Control Center and goes straight there.
That's it.
I currently use Knob and would have to say that it is the cat's meow. It would be useful to have it bring up kmix on right click. Also, it should be on default installed and used in starters config.
It'd also be nice to set the volume in kcontrol, since that's where a lot of people initially look. Of course, the nobody can find anything in kcontrol these days, so a streamlined and search-oriented kcontrol in kde4, as well as a general pairing down on certain panels (my BS.c in CS doesn't help me any in the Crypto panel) should help there ;)
knob is a nice little applet. It is simple (use left button to adjust, middle mouse button mutes). If it was installed as part of the default (think in terms of the clock being there), then usrs have a very simple quick way to check for sound.
Great work Sebast and others! K3B is getting the props it deserves! Keep up the good work!
indeed,
This, at least, is one package not many people will complain about concerning poor usability.
Superb.
Overall, there's some really useful tidbits in this report. And this part was pretty funny:
+ Explored KWrite
+ "So this is a text editor on steroids?"
Yikes, good thing he didn't open Kate! (or emacs ;)
Usually, when a student tells me vi is too hard[1], I pause the class for a minute, start emacs, and show them M-x tetris.
Then noone complains anymore.
[1] No, I don´t like vi, either, but it´s a linux admin course, and vi is the only editor kinda guaranteed to be around in case of emergency or over ssh.
Show them tetris for Vim :)
BTW. I am writing this from small intenet cafe in Hvar (Croatia) on KDE (3.1, Mozilla as primary www client).
m.
More or less all users (in this article) report the same difficulties in using konqueror and kmail. Users have been asking for more usability for quite some time now. But there has been only diffident changes in recent releases albeit major improvements would be achievable very easily. Why?
The problem is that it is difficult for the developers to sort out when a user ask for a better usability if this is just this user's problem or if it is a generic usability problem.
Many developers are also more interested in developing advanced features than simply brushing up usability. One reason is that it is difficult to know what 'better usability' means.
> The problem is that it is difficult for the developers to sort out when a
> user ask for a better usability if this is just this user's problem or if it
> is a generic usability problem.
Let's take some more or less simple examples:
Location vs File: kick Location take File, at least make it consistent (i.e. KMail->File, Konqueror->Location)
right now we have 4 (four) Configure entries in konquerors Settings menu as I just look at it, and 6 (six) Configure entries in kmail Settings, besides the unhappy quick-access settings under the Tools menu in konqueror.
remove the plugin integration from konquerors toolbar by default
use nice anti-aliasing fonts by default
remove unnecessary icons from kicker
make the kmenu understandable (freedesktop)
let all cookies be treated as session cookies by default and never ask again
rework the kwallet including its name, better disable it by default
disable the klipper by default, but still sync X11 clippboard
use plastik by default
add basic configure wizard for kmail (sending, receiving, identity), there are already these nice virus and spam wizards, the base wizard shouldn't be a problem
also add a basic configure wizard for konqueror (cookies, proxy, homepage)
add kmail's and konqueror's basic wizards to the first-time-visiting-kde wizard
disable kde tips by default
make the top left first icon on you desktop a link to your homepage
separate homepage and $HOME
... well look at the original report and add more.
Some of the above definitely is arguable. But most of that stuff is realy a simple no-brainer, isn't it.
> Many developers are also more interested in developing advanced features than simply brushing up usability.
People will love both The Usability Man _and_ The Translucency Man.
> separate homepage and $HOME
i heard thisone a lot...
maybe $HOME should have an icon with a 'folder and a house', and the homepage a 'house and a globe'.
I also find the icon for ~/Documents (which is a 'document' right now in the defualt icon theme) a bit confusing, i whould prefer a 'folder and a document' there. Actually i think every icon for a sepcific folder should have a 'folder' on it.
> I also find the icon for ~/Documents (which is a 'document' right now in the defualt icon theme) a bit confusing, i whould prefer a 'folder and a document' there. Actually i think every icon for a sepcific folder should have a 'folder' on it.
That is actually very important! Anything which is a folder should be recognisable as such. Using the document icon for the ~/Documents folder is really stupid.
My preferred solution may not be the easiest to implement, but I think it would be very effective: use a semi-dynamic icon made out of two parts - the normal folder as the base, and some other relevant icon, reduced in size and blended in in front. Normal folders would only consist of the back part (the folder icon from the current icon set). This is flexible and extensible; you don't need to create a separate hybrid icon for each special folder.
If you select the option to have folders reflect their contents, it uses static icons that look much the same as what I described, and I like the way it looks.
-- Steve
> Location vs File: kick Location take File, at least make it consistent
Why? For example Konsole has "Session" as first menu entry. "File" doesn't make any sense there.
> right now we have 4 (four) Configure entries in konquerors Settings menu as I just look at it, and 6 (six) Configure entries in kmail Settings
Who is disturbed by this? It allows quick access to clearly separated always identical (except "Configure App...") configuration modules.
> remove the plugin integration from konquerors toolbar by default
This is maybe now an option with the new Konqueror extension manager.
> use nice anti-aliasing fonts by default
On every (including slow ones) machine? It's activated within KPersonalizer for the three fastest CPU settings.
> remove unnecessary icons from kicker
Define unnecessary. Everyone has another opinion about this.
> make the kmenu understandable (freedesktop)
What are you talking about? Descriptive names are default since KDE 3.2.
> rework the kwallet including its name, better disable it by default
Where is "kwallet" used in the GUI? How is the user supposed to learn about when turned off?
> disable the klipper by default, but still sync X11 clippboard
Klipper *is* what does the syncing. What disturbs you? Do you talk about activated actions?
> use plastik by default
Already changed.
> disable kde tips by default
How will users discover/read it then?
> separate homepage and $HOME
Work on this is happening.
> ... well look at the original report and add more
Oh, you assume that we can read ourself? Thanks.
> Some of the above definitely is arguable.
But as you see almost every of your listed points has also a negatives side or requires work before it can be changed.
Some notes:
1) The problem with the "configure " items is that their similarity and arrangement suggests that they are all on equal-used. That's almost never true. Most users won't define custom keyboard shortcuts, or custom toolbar layouts, and pretty much anything useful will be under the "configure " entry. I'd hate to see the other ones taken away, but being hidden away in an advanced tab would be fine. Does anybody change their keyboard shortcuts and toolbar layout often enough to require immediate access to the dialogs?
2) With regards to the CPU settings in KPersonalizer, the thing itself is useless. It's a fundementally ridiculous concept to try to control many unrelated binary options with a single linear scale! Instead, there should four or five entries ("Use Animations," "Use Font Smoothing," etc) that the user can enable or disable. If the user really wants more control, that's what KControl is for.
3) With regards to unnecessary icons: opinions may differ, but somebody has to make the decision about what is unnecessary *for KDE*. You cannot just take the union of all options just because people can't agree on a common set. Instead, if there is disagreement, you should take the intersection of all sets, and let people add their pet icons themselves. To make an analogy to the coding world --- CORBA is what happened when people tried the former approach, and POSIX is what happened when people tried the latter approach. I'll let you interpret that how you wish.
4) With regards to ktips: if the user doesn't want to know about the desktop's advanced features, it's not the designer's obligation to tell them about it. KTips is annoying, and is a terrible interface for browsing information. There is a reason such things went out of vogue in the Windows and Mac world. If you really want to expose users to the power of KDE, put a "Learn More" entry in the application's help menu, where the cool features off the app are highlighted. If such a design is used consistently among apps, users will discover this capability, and know where to look for advanced features if they want to learn about them.
> It's a fundementally ridiculous concept to try to control many unrelated binary options with a single linear scale
That's why the check boxes exist to refine your selection/the proposal.
The problem is that there is no logical way of equating a position on the scale with a set of options it will enable. Quick: is font-smoothing a more expensive feature than opaque resize? Which does the slider enable first? Was that ordering backed up by benchmarks, or did the developer pick it out of a hat? If it's the latter, then that's just lying to the user! Does the poor user have any hope of knowing what the heck moving the bar to a given place will do?
The only hope the user has of some sort of control (remember: always make the user feel in control), is to use the manual checkbox. Well, there are too many checkboxes for the current dialog to be very nice for new users. Hence my proposal to replace the many checkboxes with a few, each enabling/disabling one general category of features. If the user really does want animations, but not opaque resize, well, KControl is there.
PS> The Mandrake installer used to have something like this too. At least their's was measured on a concrete scale --- number of megabytes of packages to install. It was still a bad idea, and I don't remember seeing it in the 10.x installer.
> Hence my proposal to replace the many checkboxes with a few, each
> enabling/disabling one general category of features. If the user
> really does want animations, but not opaque resize, well, KControl
> is there.
How is animation related to opaque resize? The former is a kind of feedback happening after the user did something while the latter lets the user tweak the interface in real time. It's easy to state that a simplification like the "eye candy vs. performance" bar is useless, but this is true with nearly all grouping of different global features for the sake of hiding the underlying complexity.
Hmm, you're right about opaque-resize vs animations. Because of that, opaque resize shouldn't even be an option in KPersonalizer, because nearly everybody is used to having it on, and if you're not, there is always KControl.
With regards to the "eye-candy/performance" bar, it's not useless because it's oversimplified, it's useless because it's illogical. A slider operates on the principle that there is a predictable ordering of values. When the user moves the slider, they know exactly what to expect. There is no such predictable ordering here. For example:
Is font-smoothing higher on the eye-candy/performance scale than opaque resizing? That's like asking if apples are higher on the fruity/fattening scale than oranges. In both situations, the first critereon (eye-candy/fruity), is subjective and impossible to order, while the second criteron (performance/fattening), is, while quantifiable, not obviously predictable.
The KPersonalizer slider is an abuse of the slider widget. It's like using scrollbars instead of sliders for volume control.
I fully agree that opaque resize should be always on (my AMD K6-2 300mHz can hanfel opaque resizing with no issues, and for VNC-like remote control one can go to KControl or use the more efficient NX protocol instead). Actually I'm annoyed about several widgets within KDE not following it setting (eg. KControl's splitter), will have to make a list of them and look what's the cause.
As for the eye-candy vs. performance slider I always assumed that it's sorted by what's actually using more and less CPU in which case it's pretty fine with me. After all it does what it propagates, allowing you to 'balance' performance against eye candy. I bet depending on personal preference most people end up putting the slider all way to either performance or eye candy anyway. =P
> > Location vs File: kick Location take File, at least make it consistent
> Why? For example Konsole has "Session" as first menu entry. "File" doesn't make any sense there.
Har, got ya. Konsole is another candidate of wired menus. Let's see how things work currently:
save a session via Settings->Save Sessions Profile
load a session via Session->MySavedSession (all saved sessions are listed in the Sessions menu; as I have lots of them they fill up to the ceiling)
new session via Session->New Shell
You see that things arn't straight here also. Better would be:
File->New Session, File->Save Session, File->Load Session
and suddenly konsole is more in line with common guis. Just my thoughts...
> > right now we have 4 (four) Configure entries in konquerors Settings menu as I just look at it, and 6 (six) Configure entries in kmail Settings
> > Who is disturbed by this? It allows quick access
Well, I personaly am not cus I got used to that over the years. But lots of new users are. Once the Configure Konqueror window is open you get confused that you are not able to configure the shortcuts, spell checker, toolbars or any other thing related to konqueror within there. You just opened the Configure Konqueror window, why shoud you close the Configure Konqueror window now and take a closer look at the Settings menu?
> > use nice anti-aliasing fonts by default
> It's activated within KPersonalizer for the three fastest CPU settings.
Why must I activate menu shadows and animation to get font antialiasing? Nobody opens the details menu to waste even more tweaking KDE.
> Oh, you assume that we can read ourself? Thanks.
You are welcome.
> Some of the above definitely is arguable.
> But as you see almost every ... requires work before it can be changed.
Hm.
> File->New Session, File->Save Session, File->Load Session
... and no menu to load a profile. Lets add File->Load Profile
> You just opened the Configure Konqueror window, why shoud you close the Configure Konqueror window now and take a closer look at the Settings menu?
You take a look at the settings menu before opening the "Configure Konqueror..." window. And you see, assumed you read top to bottom, the "Configure Shortcuts..." and "Configure Toolbars..." items before you find the "Configure Konqueror...". If a user is such short minded to forget what he read seconds before, then no GUI can help him.
So many websites have marginal JavaScript on them (most notably Internet Banking sites, but all kinds), and Konqueror just doesn't work on them. A bit of a push to bring it up to at least Mozilla's standards of JavaScript robustness and I'd never have to start a Gecko-based browser again (nothing wrong with them, but I use Konqeror for practically everything else and starting a new browser is always a small hassle, my goodness I'm getting spoilt).
KMail's fine - almost overwhelming for features and they all seem to work for me - but KNode needs a bit of polishing. Large posts get *very* slow to edit and message threading's not always up to par. I'd like to see a few improvements like "write thread/selected messages to disk as individual files/mailbox" and KMail integrated into KNode as mail sender. I wouldn't use post-HTML-article but I'm sure that if it was there, others would.
We want the formula for converting plumber into gold. There is no way for the new user to know the system in 1 minute. That's simply imposible. KDE has many details that can be better in "usability", but many of the points I see in the linked article hit the same wall. Beyond a line "usability" changes to "bad engineering", and that's something almost all FS/distro developers never are going to do. Let's se some stupids and good points:
*"It doesn't show me that I'm typing in my password?"
-> Showing the pasword length is considered a security fault in many books, for example Tanenbaun's "Modern Operating Systems"
*Thought web browser icon was not intuitive
-> What browser icon is intuitive? the solution is simple. Put a konqueror icon on the desktop with label "Browse the Web"
*"Why do I want to know about this KDE wallet system? I just want to login."
-> What the hell is security? I click in all .exe I receive, and I have a post it on mi credit card with the PIN.
*Did not understand downloading status/progress of downloading files
-> 50% is to difficult for me, should be "half the way to 0wnz the filez"
*Download manager did not give status of download
-> kget? really not?
*KDE has too much setup and installed by default. I'd rather be able to easily install what I want without having everything in the way. But the tea maker can stay.
-> That's it! then people will say kde comes with no apps. What the ...! he thinks tea maker is the best app!
There are some points he says great comments, but another...
I forget this, is really good:
*"Save Link as... does that save the link or the actual file?"
-> Why is the word "link" in the menu entry?
You can't do a stupid-safe gui, they have too much imagination.
For the Save Link as, I know what it does only because I tried.
"Save linked $object" might be friendlier, no ?
But generaly, I think the problems come from packaging. It could be clever to detect the type of machine before installing the vaio-specific modules. But the that _is_ outside the scope of KDE.
until it integrates with HAL, that is...
To tell the truth, it's the fault of the designers on this one. In English, "save link as" really does mean save the link itself. Techies can probably infer what it really means, but it's factually incorrent nonetheless. It should be "save target as," since your saving the target of the link, not the link itself.
Yes, I'm with you. "Save target as..." would be better.
"Save link as" really is unintuitive. I wondered the same thing myself. You're not saving the link, you're saving the target of the link.
If I have a web page, I can put a link to www.cnn.com on it. If I choose "save link as", it saves the HTML page on the other end of the link, not the link!
"Showing the pasword length is considered a security fault in many books"
That's security by obscurity. An attacker can always listen to how many keypresses you type or simply look at your hands. If you really care about security you shouldn't be entering passwords when someone else is around at all. The asterixes are not about security, they are about user-feedback : "hey I am really typing in a password, let's pay double attention".
I'm not convinced that vanilla KDE defaults to "don't show stars". At least the KDM kcm defaults to "one star".
Yea, this perplexed me too, I'm pretty certian that by default is shows dots when typing passwords. I don't use slackware, but I'm somewhat guessing that it wasn't kdm being used as the login manager.
Yeah! you can hear the clicks, and look the hands. But that's more dificult than looking the screen (for example in the table from behind in a college lab)
"*"It doesn't show me that I'm typing in my password?"
-> Showing the pasword length is considered a security fault in many books, for example Tanenbaun's "Modern Operating Systems""
Why not do it like it's done in Lotus Notes? It shows at random 1-4 asterisks at each keypress. That way onlookers can't guess the length of the password by looking at the number of asterisks.
Lotus notes is the #1 example of bad interface design.
http://digilander.libero.it/chiediloapippo/Engineering/iarchitect/lotus.htm
Or was that a joke I did not understand?
Eleknader
Lotus Notes certainly has it's share of usability-problems (I should know, I'm using it at this very moment!). But does that mean that EVERYTHING in Lotus Notes sucks? No it does not! The password-box if perfectly OK. Yes they do mention it in the website you linked. But I think it's OK. Of course it's not perfect security, but it's more secure than displaying 1 asterisk at every keypress.
Hi,
with that solution I am sure, people would be confused and think they mistyped, hit backspace and believe their keyboard is broken. I have seen this happen with the (existing!) 3 stars constant per character option for passwords.
The point is, if you know what's going on, you are (almost) fine. I would suggest to place a text explaining the invisibility of the password typed in case people enabled that. That would explain things.
But kdm is out of the scope of KDE anyway, more or less, isn't it? I think that is a strawman or Slackware default problem anyway. I doubt that Suse, Debian (that I would know) or Redhat (at least of 9 and RHEL 3 I know) ever did this by default, why should they? The argument about obscurity is well placed. If you can count the stars, you can also see my fingers, can't you?
For many other problems, couldn't we just let people decide themselves how to handle their security on one more personalizer step? This approach seems to be very well received. I have seen people really love that approach much. Without too many details (or just as much you like to) you can achieve shifts in the default.
We sure could make profiles that already provide defaults for cookie handling, password storage, HTML Email image downloading, etc. that avoided the need to change much things at all for many.
Yours, Kay
PS: I e.g. use KWallet at home and work as a perfectly safe harbour for my passwords. I would even like it to be possible to never enter the wallet password in the first place. That is because, at work I have only passwords that are not my own in there and I love to forget about them. And at home, I have important things there, but nobody can access KDE there, or if he can (having broken in from the internet), well, he can spy on me with keylogging anyway.
Just show the stars and forget about it! Odds are, if someone can see the number of stars showing up, they can just as easily see or hear your keyboard presses. The trivial amount of "security" that is obtained from this is vastly outweighted by the user feedback of:
Is this thing actually typing? Why can't I see anything?
and
Umm...I mistyped...or did I? How come there are 6 stars on the screen for 2 characters?
(I just installed KDE 3.3 on my Slackware system a couple weeks ago and had all four members of my family ask one of those questions over the last week. Hell, it confused ME for a short while. Yes, it defaulted to not showing any stars.
-Charles
That could also be interpreted as "Uhh, my keyboard is broken".
No, maybe a 'blinking' star/ dot/ whatever would be best, blinking each time you press a button - just to make sure the user get's a 'key down' acknowledgement...
The problem with many usability studies, even professionally conducted ones, is the complete lack of scientific method, and in some cases even logic. Why is everything based on anecdote? Where are the control groups? Why initial use of the software assumed to be representative of normal use? Where are the metrics?
It is begining to look to me like to many "experts" first use _is_ usability. You can't discount the fact that if someone (and it seems like all testers have attention deficit disorder) gets frustrated in 30 seconds they may never use the program again. Clearly first impressions are important... however they had to use something and they almost certainly had to use something they did not like because it was either mandated or the only option. This means being able to work with a horrible program is strangely acceptable for many users. Beyond that there is the devil you know. People will cling to something that can get the job done. Here's the worst part... the more their current tools grate on them the less willing they will be to accept any hurdles learning something new.
The ironies of all this remain. First of all software has some inherent complexity related to the task it performs. Second, and critically important, as crucial as a first impression is, first impressions are a miniscule part of the over all user experience. Why ignore this? I think most people know how to drive a car. Many car reviews are revisited to see how the vehicle performs and what impression it makes on the driver after a number of miles. I think two weeks and six months are good benchmark times to see how a user views software. Primarily because there is an inevitable learning curve. If you eliminate all learning you eliminate any enhanced value.
Sadly the ultimate result of first impression software is software that introduces no innovation and offers little ability to go beyond what you first did. Why do we like the web while most people don't like media advertising? 30 second sound bites don't have room for content, just cliches. While I don't want to discount the value of making a good first impression, I also don't want to devolve to where we make software only for people with amnesia. Isn't it strange that people always seem to turn to their children to help them with computers? At what age does learning and exploration become an annoyance?
I think of looking at a spreadsheet where one person has not used them and the other uses them all the time. Kspread would be too confusing to one and probably unsatisfactorily too simple to the other. The most important thing about usability remains the user, and they will never be all alike. They will also all hopefully use the program more than a few minutes.
To me usability is how well and how efficiently a program performs a task on an ongoing basis. The critique comparison in the aticle between those who just want to work and those who tinker ignores the fact that the guy who tinkers on a rocket car is going to pass the guy in the econo-box real fast. It doesn't take long to get things how I want them and them productivity reveals how usable it is. True usability cannot be discovered in a few seconds.
"Sadly the ultimate result of first impression software is software that introduces no innovation and offers little ability to go beyond what you first did. Why do we like the web while most people don't like media advertising? 30 second sound bites don't have room for content, just cliches. While I don't want to discount the value of making a good first impression, I also don't want to devolve to where we make software only for people with amnesia. Isn't it strange that people always seem to turn to their children to help them with computers? At what age does learning and exploration become an annoyance?"
Yet another great quote, thanks. ;)
"Sadly the ultimate result of first impression software is software that introduces no innovation and offers little ability to go beyond what you first did. Why do we like the web while most people don't like media advertising? 30 second sound bites don't have room for content, just cliches. While I don't want to discount the value of making a good first impression, I also don't want to devolve to where we make software only for people with amnesia. Isn't it strange that people always seem to turn to their children to help them with computers? At what age does learning and exploration become an annoyance?"
That is a strawman argument - what is at issue is not "first impressions" but making things usable from the start. You have a serious attitude problem - the users reactions are in you face, but you choose to ignore it and live in denial. You are only hurting yourself with this attitude because, if you are not prepared to listen to your users, your work will be less used by them than it could have been.
"You can't discount the fact that if someone gets frustrated in 30 seconds they may never use the program again."
In the first thirty seconds of trying SuSE, my friend gave up on Linux and Unix forever. He said it was way to hard for people. It made him log in with an account and password. That was his stated reason. Is it worth dumbing down the software so that we could get him as a user? Or could we possibly live without his use of Linux? I think the latter is more likely.
We should be focusing on true usability, and not anecdotes. The goal should be ease of use and not mere simplicity. While we do want to make a good first impression, the continuing impressions of the experienced users are much more important.
>*"It doesn't show me that I'm typing in my password?"
>-> Showing the pasword length is considered a security fault in many books, for example Tanenbaun's "Modern Operating Systems"
OK,there is a tradeoff between users expectations and security.
An easy way to improve security is to increase the minimum length of password, this way showing a star for each character won't create a too big security problem..
What on earth does your response to the KWallet observation have to do with the actual observation?!?
When I log on to a web site, I type my username and password. I do not expect to, nor do I want to, launch another application that I just have to dismiss simply by logging in. I have my passwords memorized, thank you very much. Saving them to disk, even with a master password, is LESS secure. And last time I checked, clicking on an EXE file does nothing on Linux--especially on a SPARC.
I think it's plainly obvious that KWallet is a little too in-your-face, mostly by assuming that you actually want to use it. Perhaps a "Do you want to save your passwords? Yes/No" would be better than a "Click Next to begin setting up KWallet". Just a suggestion.
You can choose not to use kwallet. And when you are registered in several pages, some of them you visit 1 time in 3 months, kwallet is a great help. That if you don't like to use same login/pass for all them. I don't have _all_ my passwords memorized, thank you very much.