eWeek: Sometimes, More Is More

eWeek's Jason Brooks gives us his take on the discussion about the complexity of KDE and GNOME in his article "Sometimes, More Is More". "I didn't switch from Windows to Linux on my home and work systems because Windows wasn't easy. I switched because as I came into contact with OS alternatives, I became frustrated with lack of flexibility I found in Windows."
Do you agree? Are we (roughly) on the right course, or should we be trimming away options?


"Because these options are fast accessible - but maybe you should first explain why before we handle why-not?"

yes you are right, that what i'm trying to do

"Anyway. Let me explain a little:

A context menu pops up with the mouse pointer being at the top-left corner.

That means the first option will be right at the mouse pointer, the second a little more farther away, etc. etc.
Also a new user will scan the options from the top, so no matter if there are 5, 10 or 300 options, if his feature is the third, he will find it in an equal time (* exception - see below)

More options don't disturb the first options."

Well i disagree on this, i think this issue is more complicated than that. There are various elements that will affect legibility, spaces between elements, icons , separators, etc. Also if we put so many options in the menus we cannot really put the most options frequently used in the top. For example, in a browser you want the first context menu to be back but you don't want forward because it's more rarely used. If you put the back option first and the forward at position 10 for example, you have a really bad and confusing behavior. If you put it with the back option you've just did the opposite of what you said. There is no easy ways to do it without removing completely options from the default.

"OK, now there is a problem when your menu is too long that it no longer fits between mouse and lower screen edge, KDE draws the menu so that the mouse pointer is at bottom-left. This is bad because your motor-memory fails."

That another problem, what resolution do you use? If a user runs in 640x480 and his menu fill the screen were back to the same point.

"Currently, on my installation, the menu works fine on 3/4 of the screen, which is OK, IMO.
With resolutions getting higher and screens bigger, I don't see a reason not to put more options into context menus."

Well we obviously disagree on this issue. A context-menu for me is just that, a menu that shows the context for the item i just clicked. I don't understand why you don't click on the main menu bar if you want so many options instead of right-clicking.

The problem is that right now we don't have choice. We have toolbar editing where we have sensible default and each has the ability to customize them. We don't have this for the context-menu and i'm not really sure about the sensible default.

For example, why do we have the first context-menu item to be "up" in Konqueror (in web browser mode)?

By Louis-Philippe ... at Thu, 2003/04/10 - 5:00am

The human brain isn't very good at memorizing complex structures without points of reference. In a 6 option menu, it's rather easy. You've got two options anchored to the top. Two in the middle. Two at the bottom. You're brain can easily hold 6 things in memory at the same time. As you use the interface, you'll start to memorize the contents of these menus, and can just select the one you want without actually reading any options or looking at any icons. This makes things very fast. Once you hit 18 entries, you overflow your mental caches, so to speak, and efficiency plummets.

Windows is the absolute worst interface to emulate with respect to interface complexity. The menus in most Office apps suck (at least they all have properties at the very bottom, though). Use something better like MacOS or BeOS for comparison. Heck, even the 3D CAD software I use for engineering classes only has 6 or 7 entries in it's context menus!

By Rayiner Hashem at Thu, 2003/04/10 - 5:00am

Maybe you guys can tell us which options should be removed from the context menus?
Since I cannot find a single unneccessary option, also I tend to add some extra scripts myself.

Make the 18 options into 6? how? bzip2 does not work on gui objects

I did not even notice them being more than 18, since the drop down menu doesn't even take up ½ of my screen.

Maybe you oughta be decreasing your font size and/or getting a higher resolution display.
I'm not gonna start looking through manuals for keyb. shortcuts to these options.

By Funk at Thu, 2003/04/24 - 5:00am

Dude. There are these little things called "keyboard shortcuts." Context menus (right-click menus) are supposed to be short and easy to scan. You should be able to digest the whole thing in once glance. If you have to scan through the damn thing to find what you want, it would be quicker to move your mouse over to the real menubar, which is better organized and easier to search. (ALT is your friend!) I have no problem with leaving "bloated menus" as an option for people who want to use the right-click menu as a replacement for shortcuts and the menubar, but useful context menus should be the default. As it stands, you can't even change the right-click menus. At least the toolbars you can customize. Besides, it's not like you're losing anything by simplifying the context menus. You're supposed to use the actual menubar for less often-used features. Everything is still available from there. It's not like you have to open the registry (yes, this is a jab at GConf, which even a long time Linux user like me finds hard to use) to change things!

By Rayiner Hashem at Thu, 2003/04/10 - 5:00am

"It has 18 items in it, and doesn't even have cut/copy/paste!"

It does if the context warrants it, like when you right click over something that is copyable, such as a file icon or selected text.

"Konqueror comes with something like 3 toolbars by default. That's far too many"

To each his own. My personal preference is 3 toolbars. One for the navigation/view buttons, one for the location bar, and one for the bookmark toolbar. Remove any one of them and my personal usability goes out the window.

"The default desktop and default taskbar is so cluttered with icons that it's almost useless."

For the panel, you may be correct. After every KDE install I end up removing the help and control panel icons. But as for the default desktop, talk to your distro! The default KDE desktop has *TWO* icons on it!

"If the KDE usability group would just take a weed-trimmer to some of the icons-bars and context menus"

Considering the plethora of icons on your desktop, perhaps you should take a weed-trimmer to your distro's packaging department instead...

By David Johnson at Thu, 2003/04/10 - 5:00am

It does if the context warrants it, like when you right click over something that is copyable, such as a file icon or selected text.
I'm running KDE 3.1.1, and I just tried to copy and post the above quote from the page. There is no copy or paste. There is a "Show Menubar" and "Copy to Root Directory" though... Copy/Paste only appears in a text box like the one I'm typing in now.

To each his own. My personal preference is 3 toolbars. One for the navigation/view buttons, one for the location bar, and one for the bookmark toolbar. Remove any one of them and my personal usability goes out the window.
Take a look at a Safari screenshoot. This is how the default setup should look. Me setup is even more streamlined. I find I have no use for the bookmarks bar (that's what ALT-B is for :) so I've got only one toolbar. Now, nobody is trying to take away your ability to customize your menu bars. You can have half a dozen toolbars if you want to. But the default should be what makes sense for most users.

Considering the plethora of icons on your desktop, perhaps you should take a weed-trimmer to your distro's packaging department instead...
Actually, I don't have any icons on my desktop. I hate desktop icons. I either use the menu, or keyboard shortcuts for common apps.

By Rayiner Hashem at Thu, 2003/04/10 - 5:00am

for example: i run mldonkey on my remote server,

and in order to get ed2k: urls qued in it for download when i click on them in konq,
i utilized a special perl script privided with mldonkey,

can you do such thing with Explorer ? (i mean, without compiler!)

By SHiFT at Wed, 2003/04/09 - 5:00am

Actually yes, as far as I know.

You need a scripting language that supports COM on Windows (almost all big ones do), and you need to implement some of the COM interfaces as defined on MSDN for parsing URLs and then retreiving data from them.

It's a very powerful interface. We're thinking of using it at work for some things.

By Mike Hearn at Wed, 2003/04/09 - 5:00am

KDE has a huge load of configuartion modules which makes admining my box a breeze!

The adminstration modules have been really useful, allowing me to configure linux without having to go in to the horrors of text files.

Gnome 2.2 is awful. You can't change the colour scheme of widgets, the button order of window frames, or be able to have a clock set in another time zone. Nautilus dosen't allow split pane view, the dialogue boxes are backwards ([no, yes], [cancel], [ok])

Gnome dosen't i18nise very well, It is just full of nasty americanisations, which isn't good for non-americans that use linux.

Overall, kde is already easy enough to use, it just needs to iron out the bugs and optimise itself a bit more. The kde 3.1.9 cvs alredy is 99% usable, and I really hope it's perfect in kde 3.!

By norman at Wed, 2003/04/09 - 5:00am

some have already touched on this, but i don't think it can be emphasized enough: there are no blanket statements regarding option complexity that if applied equally everywhere in the desktop will automatically make the desktop "better".

there are different users with different needs.
there are different types of applications with different goals, aims and audiences.
there are different tasks people undertake in different ways.
there are different technologies with different limitations and allowances.

this means that sometimes simpler is better, but that sometimes it isn't. it means that sometimes all that needs to really happen is to tweak the defaults, or reorganize the UI, or rethink some poorly thought out decisions or ...

when i hear people recommend sweeping generalizations, it makes me think that they haven't really drilled into the problem space very deep, are unwilling to, or simply fail to grasp the breadth of the challenge. when i hear people say that a certain approach works wonders for one platform (while excusing its shortcomings due to that same approach in the next breath), it makes me think that i'm glad we have choice so they can be happy too. =)

that said, there is a lot of work left to do in KDE. anyone who would like to have fun doing some work improving KDE is always welcome =)

and rather off-topic: it just occured to me how long it's been since a major service outage here at theDot. kudos to the people providing hosting and doing hte administration for this site!

By Aaron J. Seigo at Wed, 2003/04/09 - 5:00am

What makes me very nervous about this whole discussion is that somehow some 'expert' is going to solve the problem.

Experts have created the software world that exists. Where only 20-50% of the ERP installations actually work. Apple's contribution to ease of use was to make a crash into a cute 'bomb'.

Kde is almost infinitely configurable. If some wise and enterprising person wants to sell a software package based on kde, and make it less configurable, they are welcome to it. It is very easy, since the configuration manager is configurable.


By Derek Kite at Thu, 2003/04/10 - 5:00am

Complexity in KDE will not be a problem as long as its desk top is suffficiently organized that those who don't want more complex features are not forced to deal with them and those who do want them know where they can be found. Attaining this level of organization will require (1) the input of folks who we want to use the desk top, including non-geeks, and (2) good manuals. Ease of use and lots of features are not mutually exclusive.

By MH Dixon at Wed, 2003/04/09 - 5:00am

During this discussion something comes to my mind:
Anybody knows the M$ Word menus which have only the "most important"
entries when you first click and when you need more you have to click
on an arrow. IMO this definitively belongs into the UI hall of shame.
But it shows where all this discussion about hiding and removing options might lead to.
Everyone of my friends and colleagues (novices and pro's alike) turn this off
immediately after installing Office. Why? Because even novices need at least
one of those hidden functions. They don't understand why it's not in the menu.
What has this to do with the discussion? Well, even Microsoft cannot reliably
remove functions from the menu without removing at least one function which
is needed even by a novice. The problem is: What is an unnecessary function.
Not even M$ can tell after I bet hours of usability research.
We need a lot of options not because everyone needs all those options but
because everyone needs a different one. It would be even worse
than searching in a long menu to search in a shorter menu and don't find what you want!
Why are big malls successful? Wouldn't a small shop which sells bread, water and butter
of one single brand be enough? It wouldn't be so confusing. Hell, no! Nobody needs
everything a big shopping mall offers but everyone needs a single item. The key
here is not to reduce the number of goods offered but to make it as easy as possible
to find them. The dream that a computer will be as simple as a toaster will stay a
dream just because people expect from a computer to do a lot more. And they even put
the money where the complexity is: Look at Corel Draw, Photoshop and others.
They are all successful because of their vast spectrum of functions and nobody cares
if they actually need them.

By Jim at Wed, 2003/04/09 - 5:00am

Those aren't pre-defined. The office menues first popup a cache, and if you click the down arrow at the bottom, gives you the whole menu.

Those items are just the most recently used items. They go away after a long time of no use. It is a very good idea, however sometimes things expire before you want them to. In such cases, it might be nice to keep track of when people lose the item, and have to go find it again. If that happens often, the item should stay in the cache longer than it had previously. Since there is no limit to the size of this cache (other than the max size of the menu) there will not be any technical difficulties in this implementation.

The problem with office is that its settings for its cache aren't configurable, and so it comes as more of an annoyance, than a helper.

By Anonymous at Thu, 2003/04/10 - 5:00am

I think you missed the two biggest problems with the menu-autohiding:
1. A large share of the time people spent in the menus is spent for searching an item, and this is much more difficult with auto-hide menus. When you know know where the item is, auto-hide won't help you much, and
2. it may be even more difficult to find a known item if it's position can change due to the frequency of its use.

So auto-hide hurts both use cases.

By AC at Thu, 2003/04/10 - 5:00am

> Anybody knows the M$ Word menus which have only the "most important"
> entries when you first click and when you need more you have to click
> on an arrow. IMO this definitively belongs into the UI hall of shame.
> But it shows where all this discussion about hiding and removing options
> might lead to.

over my dead body. ;-)

we've discussed this exact thing several times on the usability list and it's a dead issue. these won't be coming to KDE anytime soon (probably never)

By Aaron J. Seigo at Fri, 2003/04/11 - 5:00am

That is THE most ANNOYING feature that Microsoft has EVER come up with!

By stunji at Fri, 2003/04/11 - 5:00am

I really like kde precisely because of all the options. I find that anything I want to do is easily possible. When I use gnome I find I have to really hunt for things like just making hidden files visible. (Do like extract files into subfolder though.)

I do see the argument that it appears cluttered but it seems to me that you could just make it easier to show or hide tool bars and maybe even menu entries. Then people who want a distribution thats limited can have it.

For me I want my computer to be workable for me that means I need to be able to configure it not be stuck without options.


By John Goodman at Wed, 2003/04/09 - 5:00am

Instead of trying to pull a GNOME, let's think about some quick, easy things that could be done to make things less complex, without actually removing features.

1) Make context-menus configurable. Toolbars are already configurable, so why not context menus? Both are just shortcuts for the "canonical" menubar anyway. Most of the KDE interfaces items are in XML files anyway, so this should be comparatively easy.
2) Make a standard set of "view configurations." The "super-simple" view configuration can leave all but the essentials out of context menus and toolbars. The "kitchen-sink" configuration could just copy the menubar to a context menu...
3) Make it easy to distribute view configurations. I'm thinking that many people will have similar sets of preferences. Once a particular user has their KDE setup up just as they like, they can distribute it on kdelook.org or something.
4) Get rid of "more programs" in the KMenu. Maybe it's just my distro (Gentoo) but it's a totally useless name for a catagory.
5) Take some time to nicely organize the default panel! It's just haphazard as it is, and first impressions matter. I'd suggest my setup (taskbar 75% width at top, icon bar 75% width on left, system tray, clock, and quick-access icons 75% width on bottom) but that's probably too pyschotic. How about (from left to right): KMenu, some space to differentiate it from other icons (this is currently lacking), Konqueror, Home, Help, Taskbar, System Tray, Clock. This is a pretty vanilla, mainstream setup. Most users are going to change things around anyway, but the default should at least be usable.
6) Have Konqueror take some tips from Kate about menu layout. Kate's menus are very nicely laid out, with good use of sperators.

By Rayiner Hashem at Thu, 2003/04/10 - 5:00am

Let us have all the options, and let the distro/localsysadm do the work of trimming away options they dont think the user should mangle with.

KDE is definitely on the right track by providing lot of options together with a framework for central locking down of these features. This way the responsible (be it the creater of a distro or a sysadm) can decide how the system should look.
One distro can provide a very luserfriendly environment, another can cater for the "I want my desktop to be so different nobody else knows how to use it, but I can start vi by pressing " kind of people, suit-and-tie business can adopt a very tight "only one wallpaper allowed" kind of enviroment. Everybody is happy.

By srb at Thu, 2003/04/10 - 5:00am

Although it is true that most users only use a small subset of the features of big software systems (there is that number that 80% of the users need only 20% of the functionality) we must realize that this subset is different for different users. One uses mostly list functions in his word processor to manage his club, another one does mostly frames to do character sheets for his role-playing group, another does mostly plain letters, another intergrates a lot of graphics into his documents. Different users use different feature subsets. If this is true (which I believe, because it is my own and my friends and co-workers expierience) taking away functionality is very bad. This shoud be no option. Taking away complexity while not cutting away functionality is good but very very difficult. I fear there is no easy way or standard approach to accomplish that.

By Andreas at Thu, 2003/04/10 - 5:00am

Mij ¤ 0,02 about the subject:

My opinion is that hiding options from the average user makes the desktop harder to use..
Small example which I encountered a few times this year:

A KDE user fires up Nautilus. Since Nautilus can draw the Gnome desktop, it does this also when started under KDE. Result: the user has a crippled desktop session since he does not know how to switch of the Nautilus Desktop.

However the desktop is optional and can be shut off in Nautilus, so that mr Average User can use Nautilus when running another desktop environment.

But guess what: This can't be done in the configuration of Nautilus, the user should use either a commandline string like this:
gconftool-2 --type boolean --set /apps/nautilus/preferences/show_desktop false
or install gconfig and try to figure out how to tell Nautilus not to draw the desktop.

Most Users I know delete their entier KDE and Gnome configuration, or even their Linux account, to get rid of this error, they usualy are not aware how they should configure this.

Hiding options from the GUI, and forcing users to go in to simekind of registry is not wat I call 'user friendly'. Most average users will find the desktop inflexible.

A better way in my opinion is to make all options available to the user through a consistend lineair GUI. Options that are not a lot used by average desktop users should be placed in tabs like 'advanced' etc, but not hidden somewere under the hood.


By rinse at Fri, 2003/04/11 - 5:00am

I am a new user of GNOME 2.2 and KDE 3.1. I like them both. Here are my thoughts on KDE.

* It took me two days before I found out that you can enable single-click navigation in Konqueror. I remember in KDE 2 that this was the default behaviour, and when I started using KDE 3.1 I just figured that this option had been removed. It never occured to me to look under "mouse". I only bothered to look under all the file-manager configuration screens.

* Menus. I can see the rationale for having the first menu not be called "File", but from a keyboard navigation standpoint, I always think of "alt-f" as opening that first menu. This threw me off in KDE.

* I really like Konqueror. The only thing I didn't like is the sidebar. It has too many buttons in it. It seemed daunting and confusing.

* Transparent menus. This is probably a QT or XRENDER issue, but when the image underneath the menu chages, the "transparent" menu doesn't show that change. The result of course is less than impressive.

* I really like the search tab in the kde configurator.

* The launch feedback where the launched program's icon is put near the mouse pointer is *really* ugly.

* Sound System control panel has no way of notifying the user that arts didn't get to use a realtime priority.

Do not construe these nitpicks as total criticism of KDE; on the whole I am *very* impressed with it.

By Anonymous at Fri, 2003/04/11 - 5:00am

Re: Single click

Single click is still a KDE default (unless you tell KDE to emulate Windows or such in kpersonalizer), but at least two distributions (RH, MDK) decided they'll 'help' their users by choosing double-click for them.

All good points, BTW.. And good that you found the search tab - unfortunately too many people miss it, while it can help out a lot for locating the settings one is looking for..

By Sad Eagle at Fri, 2003/04/11 - 5:00am

Ah, thanks for the info! Debian or the admins here must have set double click as the default then.

By Anonymous at Fri, 2003/04/11 - 5:00am

MDK has always used single click, and still does if YOU don't say otherwise ( a option in mdk-firsttime?).

By Morty at Fri, 2003/04/11 - 5:00am

I upgraded to Mandrake 9.1 and got double click without requesting it.
Indeed I came across this post trying to find out how to restore single click.
So we clearly have a different experience.

By Henry Izurieta at Tue, 2003/06/24 - 5:00am

I feel so connected to my kde desktop....with so many choices on how to do things... that best suits the user / task it kills me to go back to gnome / windows... where im limited.

By Mike Wolf at Fri, 2003/04/11 - 5:00am

Hi there,

I dont want to go to much into details what is better less or more, but what I want to say is that you should make a matrix for yourself of ... lets say 5 users

- general GUI user,
- mostly CLI user (admin)
- Designer
- Gamer (most or even all options ;-)

Depending if you want to get your work done or if you want to play with the settings until you get tired.

You should make some levels again. f.e. 1-5 and with the choosen usertype the options set in the level will be enabled.

I believe that Mandrake has done this in quit a nice way for the security settings.

They also used a preset of user profiles.
Beste regards

Nils Valentin

By Nils Valentin at Sat, 2003/04/12 - 5:00am

One more idea came into my mind. Generally speaking I think the most used functions should came first in a menu. So the menu should not be static but change dynamically. A software should count how often or in which interval some functions are used (save, open , copy , paste etc.). Once the software counted the frequency with which the function is used in then moves according to a made matrix up or down in a menu.

Best regards

Nils Valentin

By Nils Valentin at Sat, 2003/04/12 - 5:00am

See title. This is really all it comes down to... Make software that is flexible and powerful for users (including developers), but that doesn't include ten different options to design how a button should look like or where it should appear.
Software doesn't become more powerful by adding more options, it becomes more powerful by adding new features. Preferences and options are a neccessary evil that should be handled with care. Having certain preferences quickly available is important, so they shouldn't be buried inside of a lot of completely useless preferences. If you can solve this problem (via clever UI layout, "advanced" buttons, etc), maintain perfect stability and polish _and_ not fall for the trap to provide a "tradeoff" option instead of fixing a interface problem, then yeah, go on! I think nobody says that this would be a bad thing, KDE just isn't really there yet.
Personally I see the "think twice about every preference" mantra of Gnome more as a guideline to avoid such mistakes and to force developers to be disciplined and actually design their interface to work well, instead of basically telling the user to do it himself.

Of course you can take it too far and "dumb down" applications or simply forget about really important preferences. However, in general it should be desirerable to solve a problem instead of adding a preference to it. A small example: In Gaim <0.6, you had a preference to show button backgrounds or not. This was disabled by default which looked good in IM windows but horrible in dialogs. When enabled, it looked horrible in IM windows. Now in 0.6, this preference is gone, but instead all buttons in dialogs have backgrounds and those in the IM windows don't. Thus a preference could be removed by fixing the application. Much better! I feel that Gaim has improved a lot in usability but that doesn't mean that it was dumbed down. The opposite seems to be true, there might be more FUNCTIONALITY then ever. There are still many preferences but this time most of them (not all) are really useful and they are nicely sorted and easy to located. I was able to setup Gaim completely to my liking (which is rather special when it comes to IM) in a matter of a few minutes. Perfect!

Looking back to the article, it just shows how easily people confuse preferences with actual functionality. The author uses the accessable "show hidden files" preference in Konqueror as a good example for configurability but in fact, it has nothing to do with it. This time it's a real valuable feature to have this preference available from the main application (not buried into a preferences window). So Konqueror's way is superiour, but not because a preference was removed from Nautilus, but because of an open bug: http://bugzilla.gnome.org/show_bug.cgi?id=43472

This really needs to get fixed one day. :)

By Spark at Mon, 2003/04/14 - 5:00am