Slashdot Mirror
User Interfaces in Free Software
Havoc Pennington wrote a piece on creating good user interfaces, with special attention toward GNOME 2. See also the advogato thread.
← Back to Stories (view on slashdot.org)
← Back to Stories (view on slashdot.org)
follow the Apple UI guidelines :)
Interesting article, or set of opinions.
The major stumbling block I see in free desktop software is it's inability to innovate much further than win32 or MacOs, but there's a reason for that.
It's called familiarity - to innovate too far, would be to alienate users, so it has to be a gradual process.
KDE and Gnome have improved enormously, but they are still lacking the cohesive feel that win32 and MacOS desktops have. IOW, things like keyboard shortcuts, copy and pasting text between applications etc. are virtually universal between all different applications.
The question should be asked, are features like transparent window borders, animated icons, slide-out-menus really neccessary for a productive desktop ?
Shouldn't more development time be put into creating an efficient, robust and stable work-horse desktop and less time on the fancy bits ?
There's another aspect to this - the UI 'hobbyist' or 'tinkerer' - the very people who support and participate in the development of free UI's sometimes seem to loose the most important idea behind a good UI - the end user. Much time is spent on the idea that 'total customisation' should be the end goal - is this flawed thinking ?
How many people really want to customise thier UI to the 9th degree ? - surely the majority of people simply want a plain and effective UI that helps there productivity ?
More customisation = more code = more bugs = slower UI
A slashdotting - you get the stick first and then the carrot !
The guys at work had a chuckle at the iarchitect.com User Interface Hall of Shame. If companies like Microsoft weren't featured it wouldn't be half the fun!
Everyone enjoys a scape goat; I noticed that a lot of university professors also reference this site in their online GUI course notes!
Anyone know of any other good "chambers of GUI horrors".
Torturé par la fenêtre.
consider coffee a lubricant that helps one penetrate the coding zone
Why? Because I use Windows NT all day long at work, so that's what I'm used to. Like the qwerty keyboard, 'doze UI may not be the best, but is what most people are most familiar with. This is not a silly attempt to generate flames. I think there is some merit to just conceding the "look and feel" battle to M$ and concentrating on areas where there is a competitive advantage, like security, or just developing quality free software, with no privacy-transgressing EULAs.
Of course, tinkering with window managers and desktop setups is still a fun pastime. :)
It appears to me that a tiny percentage of all programmers know a bit about user interface, while the most of them don't have a clue. Programs that perform well but are hard to use because of an illogical interface aren't cool.
Here's a hint: before you start making the software, ask your would-be users for screenshot mockups how they would like it. You can learn a lot that way.
Didn't the Gnome Usability study done by Sun cover a lot of the shortcomings of the current GUI? It showed that the GUI was indeed created by geeks for geeks.
The report can be found here.
If something looks good and is easy to use, its faults will often be tolerated (and they can always be fixed). If it doesn't look good or it isn't easy to use it may not live long enough to get fixed.
I would imagine you use Windows then... or perhaps OSX. Because in my experience Linux on the Desktop is hardly "easy to use" or intuitive despite all the efforts made to make it so. It's fun to play around with on my dual-boot WindowsXP system but my day-in day-out uses revolve around Windows simply for the fact that everything works. I don't mind tinkering with configuration files for hours or days on end to get my Linux server to do something I want it to do. (i.e. samba, ftp server, e-mail, dns, dhcp, etc..) but when it comes to my Desktop, I just want the stuff to work. I don't find it enjoyable or even tollerable for that matter to fight my system for hours or days just to get an MPEG or DiVX music video to play back. And when it finally works, the quality and speed is sketchy at best. And the functionality of the program that I'm using to play it back with consists of "Play/Stop".
If it were just that one aspect that was a pain in the neck I could easily look past it and use Linux for my Desktop. Unfortunately I run into the same type of half-butted configuration nightmares for just about everything I want to do. Fonts, image display, games, instant messaging, web browsing, e-mail, printing, cd recording, copying images from my digital sitll camera, downloading video via firewire from my video camera, and on and on and on.
The X Window System works great when I just want to fire up some shells and ssh/telnet to a bunch of remote hosts and have them all on one screen at the same time.
That's a common point I have yet to see proven. Usually, expert users use different applications than beginners and the designer/developer of an application can adjust to that.
For applications that are being used by all kinds of users (e.g. a file manager), ease of use serves both sides. If you find something that makes a beginner's life easier, why should experts not be able to profit from that? It's not like having to throw annoying wizards all over the place was ever essential for an easy-to-use system.
The pure beginner is, IMSO, overrated. Easy to use should not mean beginnerfriendly but userfriendly. A homecomputer is not an ATM or ticket vending machine that needs to be understood immediately. If it's a little hard to learn, the user will bother once. If it's a pain in the ass to use, the user will bother every time. See the Anti-Mac Interface for details.
I've read many programmers views and opinions of UI. What they say and what they do are two different things. I mean, i'm about as qualified as any programmer to comment on UI but no matter what Havoc or anyone says about Gnome and it's usability I disagree terribly.
1. Things are as usable and only usable when people can generally agree on operation or functionality. If only 10 people agree on usability no matter how smart you believe you are, it won't be usable. The most usable applications, cars, planes, clocks, or whatever got that way through the users being able to say, "I want this and I want that". Just because you don't think it's a good idea or it will slow down performance or whatever doesn't mean you should keep those ideas out or wait to act upon them, especially simple things. This is what I see on the Gnome usability list.
2. There is no such thing as a beginner, intermediate or advanced user when it comes to usability. Sure, people need to become accustomed to a new interface but the interface should always be made so that a total newbie could walk by and get the hang of it in little to no time at all.
3. Suns usability team created CDE; have you used CDE? Was it usable to you? Ok.. I won't talk about that anymore and no offense to the Usability guys I'm sure you know more about this than I do but CDE just was not a usable product.
If you want usability in gnome I think you have to start with the basic shit. Like havoc said no one likes doing mundane work but until I'm able to drag something from Nautilus or GMC into my menu or for that matter edit my menu without being root Gnome is less usable.. It's the tiny things that count and I think that Gnome in general has neglected the tiny little things.
Would you rather jump through your window to get out of your car or use a latch mechanism to open the door?
Okay, I was thinking about this offline and I wanted to add that there's a perfect opportunity here for an OSS startup:
Give it a cool name like "SimpleFace" or something non-frightening like that (i.e. real words).
Then this company would do three things (complying to KISS):
1) Create a set of rules and guidelines for GUI applications along the lines of Apple's Human Interface guidelines. Include all of the most recent theories and practice. Publish this online. Use versions so that people can tell what's the latest draft, etc.
2) Certify apps that comply to the SimpleFace rules. Open Source Software gets certified for free. Certify non-free software for a fee. They get to put a SimpleFace smile icon on their web pages or boxes.
3) Create a set of classes - both online and corporate training - based on the guidelines. Some for free, others for a fee.
Once momentum started building on something like this, corporations would be more willing to switch to OSS software if they knew that training was going to be minimized because the apps that use the SimpleFace guidelines would be easy to use for those already familiar with other SimpleFace apps.
SimpleFace could also actively participate in the other projects as a GUI testing center. Whereas the rest of the OSS crowd might not pay attention to usability and design issues, SimpleFace would be there to help out. Providing feedback, suggestions, or even app dev for those interested.
Why am I thinking "startup" and not just "movement" or "organization?" because I think that something like this is needed now before the OSS movement loses any more momentum in the UI race from companies like M$ and Apple. (Under the theory that a startup could move faster than a committee.) How many Unix heads do YOU know that are switching to Mac OSX because their GUI is awesome? Lots.
-Russ
Me
1. If you didn't noticed "post is about GNOME"
Apparently you are the one that failed to notice what the article was about. Look again. The post was about "creating good user interfaces".
2. You shouldn't be offenced, here's why. How many people is using Windows? How many are geeks? What is average user in what field? Am I resonable to say what I've said?
I fail to see your point. If there was one.
3. We've got same hobbies I guess? Network is my field of interests. But you were the one that was complaining not me.
What was I complaining about?
I've just pointed out how you wouldn't need to do that more than once. Believe me I could extend this thing further.
Wouldn't need to configure my system more than once? That's highly unlikely. I install and uninstall programs all the time. I upgrade to new hardware, update my drivers, remove old drivers, and so on. Every so often, probably twice a year on average, I like to dump my system, start with a clean hard-drive, and install everything fresh. I also have multiple computers in my house and I'm the elected computer guru in my family so I'm often called on to help with their needs. I don't really want to spend 8 hours on the phone with my mom explaing to her how to add GTKLIB to her environment so she can compile the libraries that are needed in order to compile the base system which is needed in order to compile the user interface so she can have a GUI when all she wants to do is check her e-mail.
I have always thought that the problem with GUI in linux, is because of technical trouble!
o rt / ile_management.html.
The problem is that programmers have been way too lazy when they encounter a problem. They just ignore it or code around.
Look at
http://developer.gnome.org/projects/gup/ut1_rep
How many of these problems that people finds is because of the programmers have taken the easy route?
The programmers take the word GUI very literal.
All they have to do is taking the shell features and make a button for it.
I think that the GNOME programmers easily can make a good gui, they haven't just tried yet.
I am really afraid that they will end in the other direction, and make a featureless gnome, that only a moron can use.
Bertrand Meyer wrote a very interesting book, and designed a very interesting language. When I'm trying to do what he thought was proper and important, I like the language a lot and appreciate it's elegance and design. When I'm trying to do something that he either didn't think was proper or important, I find it intolerable.
...
His design principles lead one to construct things which always work. And depend on things which always work. Good approximations are not allowed. When and where this works, it's great. When and where it doesn't
E.g.: He disallows the overloading of operators, because the type of the operand is not an always reliable method for disambiguating the meaning. And he's right, it isn't. But it works most of the time, and in Eiffel (his language) one could create separate types for degrees_Centigrade and degrees_Farenheit, and thus disambiguate those arguments. But because a float could be either, he disallows overloading.
Where he to strictly follow his own rules, he would say that because a user interface cannot be guaranteed to work for everyone, you shouldn't have one. He's not that self consistent, but that's the only thing that saves him.
Everyone should read his book. But then they should immediately try to read a file with multiple types of data items in it that have different interpretation rules depending on the content. (Say a CVS file, where some of the fields may contain internal carriage returns.)
I always pretend to know my target audience. I just know that it's a pretense. So I try to cluster the possible incompatibilities into a cluster (or at least place an easily findable mark by them). This allows me to get the first version out in a reasonable amount of time, but still enables me to go back and adapt for a more general case on an "if needed" basis. (So far I haven't needed two. Most of my applications have a very small audience [around 20 people]. But this is a "be prepared" kind of thing that I can do with minimal cost. The other would involve either massivley expensive software purchases or considerably more work.)
I think we've pushed this "anyone can grow up to be president" thing too far.
The real problem is age and how "applications" are taught in schools, the enterprise and classes.
Schools teach children to use applications specifically. No one sits down and explains to them the concept of a file and actions that apply to a file (open, save, save as, print...) or editing (copy, cut, paste, etc.. )
If the process of educating people in the realm of computer use included a more abstract view of computers and how they work, the average joe schmo wouldn't need to "relearn" every time a new UI design came out, they'd be able to reason through it.
We moved our 8 year old daughter and 14 year old son from Windows to SuSE and Gnome, respectively. With the exception of not knowing the names of applications that do what she wants, she can get around just fine because we've taught her the basics, without being specific to an OS. She knows how to manipulate files and open applications, she understands that web browsers and can use IE, Netscape, opera or Galeon with equal ease.
This ease of adaptation is partially due to commonlality of UI implementation across applications and platforms, and partially due to their education @ home, which focuses on exploration and understanding the computer rather than a specific application.
Of course, if schools/enterprises did that, M$ would lose its edge because users would no longer be frightened to death when presented with a word processing app other than Word, or a browser other than IE.
I don't have a solution, but I certainly admire the problem.
IMO the UI of KDE is getting too complex (I know it better than Gnome). So the task is to clean it up, give useable defaults and simplify it. Especially the KDE-menu and the KontrolCenter should be cleaned up.
But what should be removed? What is a good default? Let's ask the user. KDE could collect information on what is used and how the prefs are set, and send it back to the developers.
I think noone would have a problem with that as long as: The info is anonymous, only sent with explicit consent, and it is stated clearly what information is sent.
I've heard many such arguments before, but I can't help thinking that you're mixing up a good idea (simplify the menu system) with a particular implementation that you're stuck on (top-of-screen menus, a la MacOS). How about a single, context-sensitive menu accessed from a right-click with the mouse? No, wait, we've already got one of those. So, how about having an application menu and a context menu off a right-click, with subsequent right-clicks alternating between them? And so on...
While I'm not necessarily advocating any of these ideas as "better" than the top-of-screen layout, they would appear, at least superficially, to have many of the same advantages you cite in your article: reduced clutter, easy to find (always where your mouse pointer is, some eye-catching animation to make it obvious when you click?), etc. Surely what is needed is a comparison between not just the status quo and a top-of-screen system, but between many different basic ideas, to see which are more intuitive and easy to use for the guy in front of the keyboard/mouse?
If you disagree, post your argument. (-1, Overrated) isn't your personal censorship tool for views you don't like.
Right from the start I think that the free software movement was geared by geeks toward geeks,... Of course a pretty interface is excess baggage and can lead to bloatware,
The problem is not that Linux was geared to geek audiences by geeks, but that it was, from a UI and visual standpoint, geared by nobody for no one. It needs a lot of directed intelligent "gearing" and its not getting it fast enough.
... how to put this? talented in other areas, and the combined result of so many different, valiant but doomed amateur efforts is a vile stew of visual garbage which confuses new users and continues to offend the eye of the veteran user even long after they have become used to it.
Back in 1998 I wrote to RedHat suggesting that they ought to hire a design firm or a professional design team of their own to create a visual identity for the Gnome desktop project, a project for which they were the main commercial backing. A little later after their IPO, they tried to do just that, (I'm sure it was not because I told them it was a good idea, but because it was a good idea anybody could have come up with). Unfortunately, the design firm that they contacted decided after some initial talks that they didn't want to be paid in RHAT stock and the whole thing was abandoned. Now in my letter to RedHat I didn't suggest that they bloat up the OS to make things "purty" but that they get someone in there who could make their product look like it was put together by people who knew what they were doing - instead of the way it looked then and still looks like: amateurish and jerry-rigged. It's important to realize (you folks at RedHat) if it looks like crap, (and it does), then the natural tendency of non technical computer users is to assume that it's crap all the way down to the innermost workings of the OS which they can't see . Their impression of the OS (speaking of normal PC users) begins with the graphic interface. And with the exception of errors kicked back by the OS or program crashes, it pretty much ends there too. Normal users don't run benchmarks. They don't test things to see if a combination of inputs results in a bottleneck. The concepts of cooperative multitasking and preemptive multitasking aren't clear in their minds. They don't know latency but they can feel it. The desktop is the computer for them. So you have to understand (RedHat) that what you look like on the desktop is who you are to 90% of the people you might have a chance to sell Linux to.
You are:
Kludgey and difficult to understand. The Gnome file selction dialog, for example, which came straight from Motif in its plan -is horrible. Users can never keep straight whether it's files listed on the left or directories on the right. It's confusing because the visual hierarchy is present only by its absence.
Amateurish -with many many rough unfinished edges exposed to touch and view
Butt-Ugly Literally! The default colors and palettes (used by the Gnome team which you are using unmodified) have highly unfavorable associations: mud, feet, grime, waxy buildup, faeces, fungus growth.--what the fuck are you thinking??? Do you WANT normal people to recoil in disgust? Normal people want an interface that tells them how slick, smart and elegant they are for using it. Normal people are repelled by these color associations you are showing them, which though they be euphemized as "earthy" "medieval" "natural" and a typical environment for a "Gnome", or a Monty Python scatalogical sendup of Arthurian legends, or a Dungeon's and Dragons addict's smelly apartment , they are at best dingy, dirty smudgy, unmodern looking, to put things in a strictly neutral and unperjorative way. They show moreover a profound ignorance of how color works ON PRESENT DAY COMPUTER EQUIPMENT: these color "choices" fight (unsuccessfully) the inherent tendency of cathode ray monitors to present "cool" colors well and present "warm" colors badly. Not to mention fighting the predominance of "cool" in all industrial design --going back for I don't know how many years.
In short RedHat, get someone with a clue. And since you presumably don't have one yourselves get someone with a proven track record to artdirect the project whether you do it inhouse or not.
Enough history.
1) The problem isn't about "pretty".
The immense UI bloat that already plagues free software (note the present tense; it does not threaten to arrive - it's been here for a long time) is a result of too many cooks: ie, every hacker associated with a project thinks they are a UI designer, when in fact they are programmers and not UI specialists. Their ignorance on the subject of UI is compunded by the fact that there generally isn't a separate UI department whose guidelines are backed by management authority (as they would be in a commercial software venture). The excess baggage you're worried about IS HERE ALREADY. It must go.
2) The problem isn't a lack of "pretty" in the sense that no one values it, that no one out there is trying to make their UI attractive in free software programs, but rather that most of the people who're trying are
Better UI would
-simplify
> > the choking overflow of preferences would be winnowed out to a manageable level that new users could handle without being overwhelmed and advanced preferences would be made accessible for people who are ready to handle them.
-Unify
> > the lack of design criteria for icons and toolbar pictograms would be replaced by a coherent and consistent scheme which will make users more at ease with their visual surroundings in the free software be it the desktop UI or programs running on the desktop. This would look "prettier" than what we have at present but not because it's more eye candy, nor because of any addition but it would be "prettier" because it would represent a massive subtraction.
As any graphic designer would tell you, it's what you take away from the visual experience that makes "design" from illustration or mere doodling. So, for example, in desktop icons, if they were properly designed you would actually see "fewer" pretty colors than under the no-design standard we live with currently. The reduction of usuable colors for system or program icons to an set of "acceptable palettes" may seem like a harsh measure, but this would only be SOP for any designed solution. Which is why one of Pennington's references (Matt somebody) said on his webpage that the whole thing would work better if just ONE person were doing it. One designer -- and I mean someone who designs for a living not a hobby-- would naturally tend to invent these restrictive rules for himself. He is trying to design a SYSTEM not doodling whatever comes to mind on a case by case basis. And having just one set of rules, one system, would result in a comprehensible ORDER for the user to grasp and begin to use for his own convenience. Oh yeah- I just about can hear that chorus of smarmy assholes starting to scream about "choice" now. But there's no more positive moral value to electing to have visual chaos than there is in having un-corrected buffer overflows in C programs persist because no one will get off their ass to fix it. Good design requires a difficult left brain process of logical systematization applied to a nearly intractable right brain object of visual experience. Not everyone can program and not everyone can design; in either case, those than can should do, and those that can't should shut up and be thankful that there are people around who can do what they cannot. Design is a painful process that inherently means that a multitude of possibilities are rejected in favor of a choice. Without the idea of selection, the word choice has no meaning. Having a multitude of visual choices means that no one's choice actually works; but all are democratically equal in their brokeness and futility.
It's only by having some set of rules,that drawings and colors can become at all meaningful to the user who is attempting to understand and use the computer system with fluency. Pick Designer A's or Designer B's or C's - it's not so important which of these basically competent solutions you choose as long as they are internally consistent. If the consistency isn't there, then the conveyance of meaning miscarries and the user experience is a jagged and jumbled set of interface discontinuties to which he must adapt by supplying mental workarounds and by rote memorization and every new part of his system igets mapped into his world in an idiosyncratic manner which he could never explain to someone sitting next to him or on the other end of a helpdesk phone line.
Better UI for free software requires mainly eliminating bloat and eliminating chaos, not adding things to the UI.