Slashdot Mirror
Software Usability As A Technical Problem
An anonymous reader writes "Let's face it. Poor user interface design is a big problem in software today, particularly in the Open Source world. A recent article on NewsForge addresses this problem from the perspective that software usability is a technical issue that Open Source developers can and should face and conquer, just as we have conquered other technical problems that have stood in our way." (Slashdot and NewsForge are both part of OSDN.)
Isn't this synonymous for saying that the market for computer software has grown so much that all sorts of people are using it?
I are winner
Software should be designed, not just coded. And interface must be part of the design.
Personally, i like to ask the users what they want to see. Let *them* draw the screens, then merely implement it. A three-tiered approach is best, where called for. The backend should be design and implemented according toi a decent set of guidelines and rules, and the frontend should be completely designed by the user. The middle teir is where the magic should happen, even using a nasty hack here and there.
Ultimately the disparity between those who code software, and those who use software is a big problem. Perhaps a recognition of the separate group will go a long way.
Have you read my journal today?
Open Source stuff could leverage that familiarity by create exactly the same sort of interface with all the advantages and disadvantages it provides because that would at least be familiar to the Joe Average user.
So what you're saying is that because the Windows UI sucks, it is O.K. for anything else to suck as much, too?
/dale
Why is it that the first reaction of some people here is to make an excuse?
A UI that is intuitive to navigate is getting more and more important. The reason why Windows is and has been the way it has been since its conception is tha commercial companies don't like to rock the boat. I'm sure MS has come up with tons of ways to improve the Windows UI, but implementing these changes may in their eyes, upset too many customers who are used to it. I still remember certain people getting upitty when the taskbar and Start button were added in Windows 95.
Free software, OTOH, has quite a bit more maneuvering room in this area.
For GUI applications, the UI layout is can no-longer be considered by programmers as the sole kindom of the {Photoshop|GIMP} guy sitting over there and pass all worry of it on to him or her. Just as programmers want good APIs in their code, the Human -> Computer "API" is just as critical to good and satisfactory program and user function.
See KJofol/Winamp3, and Trillian among others. You've got dozens of very beautiful skins for your apps that are a bitch to actually use. Visual beauty while nice does not a usable app make.
What is needed is a consistent, predictable interface across all of a desktop's apps. In practice, this is a lot harder than just making it look pretty.
The author alludes to the real problem with usability and open source when he comment about egotistical mailings on the newsgroups.
Too many open source developpers think of themselves as GUI experts. Until developpers are prepared to give up their egos and admit that while they may be shit-hot kernel coders, they know jack-shit about GUI development, open source will be stuck with poor usability. Unfortunately everyone seems to have their opinion on GUI development, and somehow believe that their opinion is right, despite having no training whatsoever in usuability engineering (does this remind you of how everyone is a 'pop psychologist', and a 'pop computer language expert'? -- it should).
Until developpers understand that GUI development is hard, and that it's also a science with reputable metrics, and until GUI developpers are put on the same footing as other developpers in projects, open source will continue to have poor usability.
Apologies for being so harsh on the open source world, but that's the reality of it, and we need to face that fact.
Your average linux-using developer thinks that everyone else is as smart as he is, and that command line interfaces are a good thing. The GUI is seen as a fisher-price interface for retards.
We need to get rid of this way of thinking. Software should be like a vending machine. You press a button, and it does exactly what its supposed to.
Linix and Windoze have set back the cause of usable software about 20 years!
Bollocks.
Skinning has done more to ruin usability of applications than anything else the last 10 years. Skinning has absolutely nothing to do with usability, it's purely visual customization.
Throwing out the menu/window paradigm is a very bad idea, as you get rid of the only thing the user will be able to re-use from other applications in yours.
I haven't read the article yet (on my way there now), but the parent poster has no idea what constitutes good UI, and shouldn't be modded up. I assume the article has more sane advice.
And yes, IAAID (I am an interaction designer).
Why? Because then to operate the machine, each of the users hands had to hold down a separate button making it nearly impossible for the user to inadvertently reach into the machine while it was running.
At first I thought it was a silly thing to do that would insult the operator's intelligence (who would be stupid enough to reach into a compactor while it was running?) But one of the operators confided that it was a great idea because after being burned out from working a couple of double shift days in a row, he didn't want to loose his hand from a simple operational oversight.
The operational interface was well recieved because we gave the users ownership in the design process. I think that the same should apply in designing software UIs.
No no no!
Artists give us interfaces like ATI's TV recording software. All flash and no function. The more artistic freedom an app gives to skin designers, the more time I have to spend squinting at the cryptic emblems and trying to click on the 3-pixel-wide "play" button. Look at an old version of Windows media player (before v6), and marvel at how much easier it is to use than WMP 9 or Winamp 5. It uses the same widgets as the rest of the desktop, so you don't have to spend any time at all trying to decide where to click to activate each button. Artists understand what looks good, but very few of them have a grasp of what's easy to use.
It's better to write everything for a standard set of GUI widgets, and provide a mechanism for theming those standard widgets to look cool. That way, all your apps look consistent, and you can change the look-and-feel without having to re-learn all the interfaces.
0 1 - just my two bits
I've tried to switch over several times, and I just can't do it. KDE and Gnome get better and better, but they are still so different from the Windows I've been using for the past 14 years. I always have the same problems: I can't find things, wizards don't work properly forcing me to go to HOWTOs and the command line/conf files, and theres not enough integration between the window managers and X. I'm quite technically competent and I get better and better with Linux everytime I try it, but for the average user or your mother/grandmother there is still so much work to be done.
No matter how shity Windows is, one thing you can't argue is it's ease of use.
Anyone from a five-year-old to a WWI veteran can sit behing a Windows PC and be browsing the Internet and checking mails in no-time. (mind you, i'm not arguing the risks of this)That is what OSS should try to learn: simplicity. Average users like it simple and straight-forward, and IMHO that's *one* of the reasons for windows success.
--Just as important, the average user is by now used to the Windows interface, and it wouldn't be that bad of an idea to give them the power and strength of OSS with a windows-like interface which they are more comfortable with.
Is it just me or are video games way ahead of other apps on user interface? These days most people can pick up a game and given the general type (fps, driving, rts, rpg) have a pretty damn good guess at the interface. It's not that the game authors have agreed on a standard interface for each genre, it's that they've figured out the things that frustrate new gamers the least so they enjoy the game more with less manual reading. When was the last time you had to read a game's manual to actually jump in and play it? I mean just the basic playing around, not the detailed stuff.
Why haven't desktops and apps incorporated advances from here? Let's take an old RTS, say Command and Conquer. The designers figured out how to make a USEABLE virual desktop that DOESN'T SUCK! You can navigate around this huge screenspace and the radar keeps track of where you are. Also, how do they handle things similar to launching apps? Well there's a sidebar full of big easy to distinguish one click icons, and a set of tabs at the top that switches what set of icons is displayed by type (units, buildings, etc). Seems pretty easy to figure out to me. Want to quickly get back to the thing you were last working on? You can designate hotkeys with ctrl+number an then pressing the number jumps back to it. Some RTS's have seperate select and change focus but all seem to use a similar hotkey system.
One of the things that keeps me happier with windows than linux is the at least moderate effort at standardized interfaces. Most apps of simlar types have similar interfaces and I don't have to relearn all the terms that someone decided to use THEIR names for. Every time I see a custom media player or something with this horrible neo-future interface on windows I cringe, because it's such a bad idea. I don't want to spend time relearning how to use a media player just so it can look cool, I want to watch media with it. On linux it seems every app suffers from this "I want to look unique" urge, or a complete lack of asthetic design whatsoever. So your choices are pretty and confusing or ugly and confusing.
Introducing the new Occam Fusion! Now with sqrt(-1) fewer blades!
When contemplating a paint or sculpture from Picasso, you may be there and think for minutes trying to understand what Picasso was thinking while doing this paint/sculpture.
So, I don't really think you really mean artists, but rather than designers. That's not quite the same.
Achille Talon
Hop!
Usable doesn't mean pretty. Pretty doesn't mean usable. Artists can add aesthetic polish, but if they don't know anything about usability, they'll just make the problem worse. Look at Kai's Power Tools or the various other applications that try to look happy or fun but end up being totally non-standard and difficult to use.
Skins are not a solution to usability. Skins are a punt. To me, skins represent everything that's wrong: the software developers doesn't feel like spending the effort to time on design and doing usability testing, so they throw on a skin system and let the user deal with it.
How many users actually go create their own customized skins? And most skins out there usually cater more to aesthetics than utility.
Plus, there's the perpetual problem where every application has its own skin, and nothing is consistent with anything else. If necessary, global themes should be used for personalization; per-application skins are a mess.
Good lord, no. Please, please don't reinvent GUI widgets. Lack of consistency is one of the problems, especially in the OSS world where there are a zillion and one widget toolkits. Do you want a dozen different textboxes where some of them allow copy/paste and some of them inexplicably don't? Or maybe some of them can't handle Unicode, or maybe some of them don't have keyboard shortcuts to select text, or who knows what else.
Standardize. Stop bickering, stop wasting time reinventing things, and then everyone can focus on real usability issues.
In an open source world everyone can customize the software to suite their needs so you sacrifice Consistency and Usability for Flexibility. Advanced users are happy but novices loose out.
If you want to improve usability in Linux or other open source projects you need to put someone in charge of the UI. Linus is the de-facto gatekeeper of the kernel but the UI seems to be fair game for just about anyone.
(A Former MS UI Guy)
Most companies I work for as a contractor consider the UI design as an afterthought, an unwanted burden, or a mere exercise for the programmer who was assigned the interface screen. The development managers have been hardnosed pragmatic guys who see no sense in spending their budget on any 'needless' items like psychology and design of a proper UI. These clowns also see no sense in developing state diagrams for the control flow on interfaces. The result is often interfaces that have unlearnably convoluted navigation. This is just unforgivably bad design practice. Sometimes I have to state chart the UI to prove that the interface is broken and bad. I often see interfaces that dynamically change their functionality - same screen, but buttons and selectors come and go depending on the state, new navigational connectivities invisibly appear and disappear - all of which confuses the hell out of users.
See, a user first encountering an interface has to build a mental model of meanings of objects, control flow/states, and navigation. Your goal as a designer or programmer is make the UI design easily learnable and usable. That's both a science and an art.
I've also seen far too many UIs employing flashy objects that interfere with the readability. I don't care if a button looks like a 3D gem, if I can't read the friggin text label quickly and easily under the gloss, it's a failure. Yet I've seen $6 million corporate software with unreadable browser-based interfaces apparently designed by a 16 year old Web designer with attention-deficit disorder.
Visual readability, learnability, ease of understanding navigation, three major rules.
The real key to Good UI design is consistency. Many open source projects have unfinished features, differing UI conventions and throw the user curve balls. This can be expected for testing and non stable releases. However any release labled stable build or 1.0 and so on should have a clear consistent UI and NO I repeat NO unfinished features.
This alone would help greatly. When a user downloads a stable build binary he should never see a menu that doesn't work or a radically different approach to a task that doesn't fit with the rest of the app. CVS snapshot builds and testing builds are a different ballgame.
Also Stable builds need to be clearly marked as such and stressed as the "polished" version.
No, I assume what he means is that if MS, with all its resources, has a hard time in the only area where they seem to make a serious effort, then it must be a difficult task.
Another issue I think needs to be discussed is the way people's biases influence UI design. Some people, especially younger users, seem to think GUI==good automatically, and thus, the more eye candy a UI has, the better it must be. Conversely, they think that a less graphical interface is automatically primitive, and that anyone who criticizes excessively flashy interfaces must be an old fart pining for the days of punch cards. Such people will see lots of eye candy and get a warm fuzzy feeling, and will think that UI is "easier to use." Even though all the stuff just gets in the way, and he has to go down through 4 or 5 levels of menus and/or screens to do the simplest thing.
Unfortunately, such people seem to dominate UI surveys, and UI designers get the message. The result, for me, is endless frustration as the UI keeps trying to "do things for me" and I keep having to hunt some setting down in the dungeons of the preferences editor somewhere to turn off yet another annoying feature.
Speaking of which, does anyone know how to tell XP to stop rearranging menus and/or hiding half of the options? That's such a PITA -- who the hell thought of such a moronic thing?
I'm a software developer, I work on commerical software. What you propose is great, lets fix the usability problems, and then worry about the technical 'behind the scenes' problems.
The problem is, that there are many cases where a seemingly minor UI change to the program would downright destroy the backend. For example (and it's a crappy one, I know, trying not to overcomplicate it...) Take a look at how the clipboard works. You can copy one item of infromation on it at once, and take one item off at once. As an end user, I don't understand why I can't copy text from notepad on it, then copy an image from paint, and then paste the text I had into Word, and paste the image I have back into paint. Obviously all the changes needed would be to know what I want to paste. Problem comes when as a programmer I now have to figure out what to paste where. Text into notepad, that's easy... but what about Word... now what, image or text? Okay, lets ask the user... oh wait, I can't, because I'm not able to do any UI in another program (hypothetically here...)
Yes it can be solved, but from a user point of view, it's minor, and from a programmers point of view it is damned complex and not worth the trouble. Let the user do two copy operations, instead of me having to write and debug thousands of lines of code that is trying to assume what the user wants (and the user will bitch about if I get it wrong anyways). Add to that most OSS developers are doing this for free, and are not going to want to rewrite their backend just for a seemingly minor UI change, which isn't going to make everyone happy, just a few people who complained.
What some people find intuitive is complex for others... there is no happy median... there will always be UI's that are not liked by some... there is no perfect UI design out there... and very few people willing to try and find it, especially for free.
---
Programming is like sex... Make one mistake and support it the rest of your life.
It doesn't. I'm an architect and I regularly observe UIs that have no sense to them whatsoever. Open Source acts usually as a meritocracy and I've never found a coder who was willing to redesign his entire application because the UI sucked. It's not a chicken and egg problem (as other posts around seem to indicate) since the UI always comes last.
I once considered starting a project that designed application interfaces for tasks that were needed in hopes that some coder would come along behind and actually write them. (I had a great idea for a clock that doubled as a date/location/world time zone applet.) But we have no influence. UI is considered like the body molding tacked on to American cars half way through a model's life to re-energize sales. It's never considered as an integral part of the design the way someone Porsche does.
There is no need to use a SlashDot sig for SEO...
Presentations went through a similar trend. Thanks to Powerpoint (mostly), the emphasis shifted from conveying the essence of your ideas simply and succinctly to making things pretty. It really was due to the laziness of audiences: if your slide had lots of colors, then it must be good -- they weren't really paying attention to it anyway.
My advisor got into that trend. It started when he told one of the grad students to add some color to the slide, which was a block diagram. "What color should I add?" was the somewhat sarcastic response. It didn't matter; any color would do, as long as it was exciting. (No, there were no differences between the blocks that could be expressed by color.) Then he had some of our undergrads do some presentations. One guy went wild, throwing in pointless animations and sound effects that did nothing except show off his Powerpoint skills. But then the advisor started encouraging this from everyone.
Fortunately, most people didn't go for the over-the-top stuff, but they still would do things like put shadows on boxes in block diagrams, even if that meant using smaller boxes, and therefore smaller fonts for the text. So who cares if the people in the back can't read your slide -- it's pretty!
Some of the best presentations I saw, BTW, were by someone who was giving an extemporaneous talk, and was drawing diagrams with a single marker on a clear sheet.
What apps and OSes need is scalable UIs - UIs that scale as the knowledge level of the user grows. A total novice, non-technical, casual users should be just as comfortable and productive as a hard-core, 80-hour-per-week developer. This has not happened yet because there are two distinct camps in UI development. Profits in the mass market drove closed source, mass-market software to create useability on the low-end. The natural interests and abilities of its contributors drove open-source to create useability at the high end.
The biggest challenge to scalability is creating inuitive metaphors or abstractions between the human interface (i/o modalities) and underlying digital constructs that does not get in the way of the power-user. Apple's early OS effort were great for the novice, but derided by more experienced users - the UI was not scalable in the upward direction. In contrast, Unix/DOS/CPM was fine for power-users, but it arcane command interface made it not scalable in the downward direction.
I suspect that the answer will be concepts like Mac OS X that combine GUI and CLI elements. But even OS X is not as scalable as one might like because it is really an intuitive Apple GUI grafted on to a separate powerful *nix CLI core. Although novice Mac users can "graduate" to the command line, the transition is not smooth -- using Finder does not teach one how to use ls, cd, mv, cp, rm, etc. Rather than being scalable in a continuous sense, Mac OS X offers interfaces at two different scales - the intuitive GUI and a separate power-user CLI.
Perhaps future OS/app UIs will be truely scalable -- early GUI use will seamlessly teach the user and help them slowly become more powerful users. Developign scalable UIs will require contributions from both novice-oriented usability experts and power-oriented developers. It will require forethought and coordination so that the disparate elements of the system are "consistent" without being inflexible.
Two wrongs don't make a right, but three lefts do.
Ten things you can do to make your program at least tolerable for end users:
Perhaps there is an unspoken rule in the community that "easy user interfaces" = "simplistic programming" and perhaps that causes one to lose points in the open source "meritocracy"?
I really like your idea of designing interfaces for tasks and then developing the code to support the interface next. It implies that the user's need is defined first by the design of the interface. This locks the programmer into coding in a way that meets the user's need exactly as specified by the UI. It's a shame that didn't take off . . . But perhaps that doesn't leave enough creative freedom for the programmers to feel the project is "fun" enough to work on.
Everybody says how difficult it is to design the proper gui. I think that if a little common sense is applied, then guis can be functional and pretty at the same time. Here are some tips:
1) don't clutter things closely together; provide the proper spacing between elements of the gui. KDE/Gnome severely violates this rule.
2) use soft colors. Harsh colors make the user tired very soon.
3) Use bitmaps and labels that have a clear meaning to the user, not to the developer.
4) Be consistent with interfaces. The File menu should be there to open/close files; Ctrl+S must save the current document etc. There are lots of established conventions that work really well.
5) group similar things together (similar by concept).
There are really good sites with lots of examples of what or not what to do. But I don't think it is extremely difficult to make a GUI. All that is needed is plain common sense. I am a programmer, but people never complained about my GUIs.
Open source is full of people that are completely out of touch with reality. The people who are involved in OSS have outright contempt for those who 'merely' use the software
This is insightful? It's pure crap. There are certainly people who wrote OSS who do have contempt for the "mere" users, just as there are plenty of people who write commercial software who have contempt for the users. In general, though, developers of end-user applications, whether OSS or commercial, feel no such thing. They want their apps to be usable, because it's really cool to have lots of people using your stuff. That doesn't mean they know how to make software usable, of course. Wanting to and knowing how are different things.
Very bad example, though. The above is fantastically usable... find me a GUI app that can accomplish the same purpose as quickly and easily. The above is an excellent demonstration of the difference between ease of learning and ease of use. The UNIX command shell is extremely powerful and easy to use, but it is not necessarily easy to learn.
An easy to use interface is one which makes it possible to accomplish the desired tasks quickly and easily, without unnecessary steps or wasted motion.
An easy to learn interface is one which allows the user to accomplish the desired tasks without training (or significant effort to figure out how). Note that this concept is fundamentally different from ease of use in that while ease of use is an absolute (for a given task set), ease of learning depends heavily upon the user's other experiences and is achieved mostly through similarity.
These two axes of ease are nearly orthogonal, although they often seem to be somewhat opposed to one another. There are plenty of examples of apps (particularly in the Windows world) that are easy to learn but hard to use, and lots (particularly in the UNIX world) that are hard to learn but easy to use, but there are also a precious few that are both easy to learn and easy to use (many of them in the Mac world, actually). And there are an unfortunate number that are both hard to learn _and_ hard to use (Easily found on any platform). I'm sure if you think about it for a moment, you can come up with examples in all four categories.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
Graphic designers generally suck at user interface design.
User interface design is wildly different from graphic design. As a matter of fact, there are probably more industrial designers that would do a better job of doing software user interface design than graphic designers.
I'd say that a lot of awful websites out there were due to people with traditional publishing and graphic design experience trying to apply old knowledge to the Web and failing.
May we never see th
The problem is, that there are many cases where a seemingly minor UI change to the program would downright destroy the backend.
And that's precisely why we usability folks advocate designing the UI at the *start* of the development process, so usability is there from the get go and programmers won't have to re-write zillions of lines of code.
Unfortunately, designing UI before writing code is seen as heresy by the Unix Culture that dominates Open Source, often being referred to as a "proprietary" development methodology. And this is one of the big reasons why Open Source usability sucks.
Ergonomica Auctorita Illico!
Very bad example, though. The above is fantastically usable... find me a GUI app that can accomplish the same purpose as quickly and easily
It's a good example, but not for the reasons you are thinking. GUIs don't do this because it's a completely uncommon task. (If anyone actually cared, it would be easy to add a "save as text" button somewhere in a filemanager.)
However, the CLI Fanclub can't get past the the idea that a GUI is crippled because it can't do the stuff nobody really wants to do anyway. They are completely confused between the concept of a "user" interface (make everyday tasks easy) and a "programmatic" interface (be infinitely flexible).
(Now someone's might come at me about how they use grep/find 300 times a day, but do they really do that more than simple directory browsing or copying random files from point A to point B?)
I think the original poster was being a little extreme, but you do get the idea that Unix Filemanagers are developed for "other people" and not for "us" or "everyone".
Business. Numbers. Money. People. Computer World.
What you're commenting on is the Unix vs Microsoft way of handling files/folders. And if your argument is intiutivity: take a look at Mac OS X, it's also based on the / structure. Some things are just more easy with this way of working. Especially mixing rw an ro partitions, network partitions, ... transparent to the user/applications. If you log in to a *nix box with home directories which are actually located on some nfs server, you'll find them under /home/youruserame. If you look for a program, you'll find it under /usr/bin. It doesn't matter if it is a rw partition on your local harddisc, or a nfs partition on some server, or even a ro cd, it's in /usr, where you expect it to be. With windows its a lot less intuitive to find My Documents on C: at box A, on D: at box B, and maybe on Y: at your work. Even worse if you want to install a program on a network disc. If you want to install it on M:\program files\, but the windows dll's are on E:\windows, you're really fucked. And nowadays most computers have more than 1 harddrive partition. In the old days A: was the floppy, C: the harddrive, D: the cdrom. But today I find computers with C, D and E are harddrive partitions. F and G are cdrom and dvd. J and I are shared music and movie drives, and M some online webspace. You call that intuitive? With the "everything is in /" structure you can have a /usr on the local disc, /usr/bin on a ro local disc, /usr/lib on a nfs share, it's just more flexible.
And to find your files, just use common sense: most Linux user mount their cd's under /mnt/cdrom, or /mnt/dvd, a floppy is usually: /mnt/floppy. Install programs in /usr: the executable goes in /usr/bin, /usr/sbin if it's a static one. Shared objects go in /usr/lib. Files belonging to a particular user go in /home/username. In that directory you can create folders for you text documents, multimedia files, ... in whatever way you want. If you want to share those files with other users on the pc, create a directory like /public.
I agree that at first the structure on a unix system looks like chaos, you seem unable to find anything in it. But when you understand some of it's basics, you see how logical and powerful it is.
Manuals is another discussion, and for every project the documentation is different. But IMHO the Gimp documentation is good.
This is one lame signature, please read the message above instead.
Now you may want to call me a heretic, a troll, and a baiter of the flame, but listen my brothers and sisters to what I am about to speak none the less.
Many have you say, "Linux isn't any harder to use than windows/mac." That my friends is a lie. Still many more of you say, "Linux is almost there!" Again, I say that is a lie. I know! I have been using Linux since 1994. It has suckethed in the past, and it sucketh today. Has it improved, yes, but it is still quite bad. While I can only speak for the state of GNOME, I can say that it is actually becoming harder to use in the name of useability. How you ask? Why the file chooser dialog has no filename entry. Support for typing filename or URIs, things that have been included in everyone of the filechoosers ever developed is hidden under arcane keystrokes and even then lack the support of 2.4. Abilites that distinguished the GNOME desktop from others have been removed in recent years inorder to make it "more intuitive", which is merely a synonym for "poorly cloned in the broken in the way of Redmond".
The linux user experience is one of confusion and inconsistency. Applications don't look the same. Applications don't behave the same. Applications having improper interface criteria ("Edit|Preferences"? Why would I look for configuration details in the same menu that I use copy, paste, and search the text in?) Installing packages leaves them unconfigured, or configured with broken defaults. Too many times, the user is forced to enter commands at the terminal, or edit cryptic configuration files. Things that should be automatic aren't.
I postulate that this situation could be be resolved with a two pronged approach. First, a distribution that doesn't try be the One True distribution with every conceivable package in it. It should have one desktop environment, one office package, one media player, one emailer, et cetra. In short, one and only one of every software type. This simplifies package configuration, and enables almost complete autoconfiguration.
Secondly, all the user applications must be tightly integrated. There shouldn't be a mixture of say Gtk, GNOME, wxWindows, and Motif applications. All applications be of the same toolkit and of the same desktop enviroment. This will help make the user experience more cohessive. Unfortunately this isn't enough either. There has been developments in some of the required software that seem to be actually detremental to the user experience. Either a new enviroment will need to be developed (*bleh*) or perferably patches against an existing enviroment/applications developed. (Think Ximian, only not based on a cult personalities.)