Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 2 years ago.
Improve this question
Pretty much every other editor that isn't a vi descendant (vim, cream, vi-emu) seems to use the emacs shortcuts (ctrl+w to delete back a word and so on)
Early software was often modal, but usability took a turn at some point, away from this style.
VI-based editors are total enigmas -- they're the only real surviving members of that order of software.
Modes are a no-no in usability and interaction design because we humans are fickle mammals who cannot be trusted to remember what mode the application is in.
If you think you are in one "mode" when you are actually in another, then all sorts of badness can ensue. What you believe to be a series of harmless keystrokes can (in the wrong mode) cause unlimited catastrophe. This is known as a "mode error".
To learn more, search for the term "modeless" (and "usability")
As mentioned in the comments below, a Modal interface in the hands of an experienced and non-fickle person can be extremely efficient.
Um... maybe there isn't much of a need for one, given that Vi/Vim is pretty much available everywhere and got the whole modal thing right? :)
I think that it's because vi (and its ilk) already occupies the ecological niche of modal editors.
The number of people who prefer modal and haven't yet been attracted to vi is probably 0, so the hypothetical vi competitor would have to be so great as to make a significant number of vi users switch. This isn't likely. The cost of switching editors is huge and the vi-s are probably already as good as modal editors go. Well, maybe a significant breakthrough could improve upon them, but I find this unlikely.
#Leon: Great answer.
#dbr: Modal editing is something that takes a while to get used to. If you were to build a new editor that fits this paradigm, how would you improve on VI/VIM/Emacs? I think that is, in part, an answer to the question. Getting it "right" is hard enough, competing agains the likes of VI/VIM/Emacs would be extremely tough -- most people who use these editors are "die hard" fans, and you'd have to give them a compelling reason to move to another editor. Those people who don't use them already are most likely going to stay in a non-modal editor. IMHO of course ;)
Modal editors have the huge advantage to touch typists that you can navigate around the screen without taking your hands off the home row. My wrists only hurt when I'm doing stuff that requires me to move my hand off the keyboard and onto the mouse or arrow keys and back constantly.
Remember that Notepad is a modal editor!
To see this, try typing E, D, I, T; now try typing Alt, E, D, I, T. In the second case the Alt key activates the "menu mode" so the results are different. :oP People seem to cope with that.
(Yes, this is a feature of Windows rather than specifically of Notepad. I think it's a bad feature because it is easy to hit Alt by mistake and I don't think you can turn it off.)
VIM and emacs make about as much user interface design sense as qwerty. We now have available modern computer optimized key layouts (see the colemak layout and the carpalx project); it's only a matter of time before someone does the same for text editors.
I believe Eclipse has Vi bindings and there is a Visual Studio plugin/extension, too (which is called Vi-Emu, or something).
It's worth noting that the vi input models survival is in part due it's adoption in the POSIX standard, so investing time in learning would mean your guarenteed to be able to work on any system complying to these standards. So, like English, theres power in ubiquity.
As far as alternatives go, I doubt an alternate model editor would survive a 30 day free trial period, so its the same reason more people drive automatics than fly jets.
Since this is a question already at odds with the "no subjective issues" mantra, allow me to face that head on in kind.
Non-Modal editing seeks to solve the problem caused by non-modal editing in the first place.
Simply put, with Modal editing I can do nearly everything without my hands leaving the keyboard, and without even tormenting my pinky with reaching for the control, or interrupting my finger placement by hunting for the arrow keys.
Reaching for mouse completely interrupts the train of thought. I have hated the intense reliance upon this with Intellij IDEA and Netbeans for many years. Even with vim-style addons.
Most of what you do has to do with fine-tuning with very small increments and changes within the same paragraph of code. Move up, move over, change character, etc., etc. These things are interrupted with control keys and arrows and mouse.
Though not really answering your question, there used to be a "modal like" way to write Japanese on cell phones before :
The first letter you hit was a conson let's say K, and then, and then the next key you would hit would have the role of a conson. (Having two conson in a row is impossible in Japanese)
Though it was main a few years ago, today it's only used by people who really want to hit fast.
I think the answer to the question is actually there are quite a few modal text editors that aren't forks of vi/vim. However they all use the vi key bindings. Vi users get the key bindings into their muscle memory so relearning a different set of key bindings would be really hard, so no-one would create a different set of key bindings.
But lots of different editors have re-implemented the vi key bindings from scratch. Just look at this question about IDEs with vi key bindings. At least half of the answers are editors built from scratch that implement vi key bindings, not versions of vi embedded.
I recently came across divascheme - an alternative set of key bindings for DrScheme. This is modal, and part of the justification is to do with RSI - specifically avoiding lots of wrist twisting to hit Ctrl-Alt-Shift-something. The coder has done an informal survey of fellow coders and found that emacs users suffered from more wrist pain than vi coders.
You can see him doing a short talk at LugRadio Live USA. (The video is a series of 5 minute talks and I can't remember how far through it is, sorry - if someone watches it and posts that here I'll edit this post to say when in the video it is).
Note I have not used divascheme.
The invention of the mouse took one mode and moved it to an input device, and context menus took another mode and moved it to a button. Ironically, the advent of touch devices has had the reverse effect, producing multi-modal interfaces:
aware multi-modal - touch and speech are aware of each other and intersect
unaware multi-modal - touch and speech are unaware of each other and conflict
The traditional WIMP interfaces have the basic premise that the information can flow in and out of the system through a single channel or an event stream. This event stream can be in the form of input (mouse, keyboard etc) where the user enters data to the system and expects feedback in the form of output (voice, vibration, visual, etc) when the system responds. But the channel maintains its singularity and can process information one source at a time. For example, in today’s interaction, the computer ignores typed information (through a keyboard) when a mouse button is depressed.
This is very much different from a multimodal interaction where the system has multiple event streams and channels and can process information coming through various input modes acting in parallel, such as those described above. For example, in an IVR system a user can either type or speak to navigate through the menu.
References
User Agent Accessibility Guidelines working group (UAWG): Keyboard Interface use cases
W3C Multimodal Standard Brings Web to More People, More Ways
Next steps for W3C work on Multimodal Standards
The Future of Interaction is Multimodal
Beyond Mouse and Keyboard: Expanding Design Considerations for Information Visualization Interactions - naturalinfovis_infovis2012.pdf
Setting the scope for light-weight Web-based applications
Jan. 26, 1983: Spreadsheet as Easy as 1-2-3
Multi-modal design: Gesture, Touch and Mobile devices...next big thing? | Experience Dynamics
Related
Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
This question does not appear to be about a specific programming problem, a software algorithm, or software tools primarily used by programmers. If you believe the question would be on-topic on another Stack Exchange site, you can leave a comment to explain where the question may be able to be answered.
Closed 8 years ago.
Improve this question
There are many alternative keyboards to the standard US keyboard layout (called QWERTY).
Some examples include Dvorak, (and variants like Programmer Dvorak), Colemak, AZERTY, Workman layout, etc.
Do any of these confer a benefit to typing speed, accuracy or hand/wrist-health?
And, if so, which one of these should I choose as a touch typist if I am regularly programming?
Short answer:
If you are happy with your keyboard layout, stick with it.
Long answer:
I will try and aim to make this as definitive and explanatory an answer as possible. To understand a bit where I am coming from, allow me to express my own journey through this jungle:
I am a computer science student who started out with the German QWERTZ keyboard, typing at about 100 WPM (words per minute). When that turned out to be horrendous for programming, I moved to QWERTY. Then, I got taken in by the hype and turned to Colemak. After mastering it, I discovered there was a layout optimized for programming, and switched to Programmer Dvorak. Finally, still not happy, I tried to design my own keyboard layout semi-scientifically. And finally, now, I am typing these lines on QWERTY. (To save others the trouble and pain I went through).
Therefore, I will try to argue in my answer both from personal experience as well as from scientific published data.
The main arguments for all the alternative keyboard layout hype can be summarized to three major points:
The QWERTY keyboard is slow and was designed to slow typists down.
Excessive use of QWERTY causes Carpal Tunnel syndrome and is bad for your health.
Dvorak/Colemak/< Insert alternative layout here > was optimized to increase speed/accuracy/health
Let's go through this one by one:
First, the argument that the QWERTY keyboard was designed to slow typists down is simply not true. It was nicely debunked in this question. The QWERTY keyboard was designed to stop the keys from a certain model of typewriter to stop jamming. Rest assured, we will discuss the "QWERTY is slow" myth in a minute.
Second, the ultimate argument that advocates of alternative keyboards love to use is that QWERTY causes Carpal Tunnel syndrome, because it strains the fingers.
What's amazing here is that this is actually an Urban legend which has persisted despite it being discredited. See this question here. To quote from the answer by Graeme Perrow: "It seems that using computers in general does not cause carpal tunnel syndrome, regardless of the type of keyboard."
Finally: If QWERTY wasn't made to slow typists down and doesn't cause any illness, why use another keyboard? The answer usually offered is because other layouts are faster and have the keys aligned in a "smarter" way.
We are told how much faster typists can be when they use Dvorak instead of QWERTY and how the home row of colemak offers great benefits to productivity and speed.
We are treated to an avalanche of impressive-looking percentages, of how much faster and accurate you can be on an alternate keyboard, rather than a humble QWERTY.
However, if you look at hard, scientific evidence, you find... nothing worth writing home about. Indeed, there are two very interesting posts here and here: It turns out that the (very hard to objectively measure) speed gains are a measly 2% to 4%.
This mimics what I myself have experienced: If you are a trained typist, then switching doesn't give much of an improvement. After I had finally finished my switch to Dvorak, I was still typing at roughly 100WPM. If you want to go beyond that, you have to type a lot during your day.
I believe that the reason people observe a speed-boost when they switch is that they have to retrain their muscle memory from scratch. Which, if they do diligently, is rewarded with a faster typing speed. The irony is: I conjecture that if they had "retrained" QWERTY from scratch, they would have obtained the same speed increase.
Additionally, my own error rate didn't go down with Dvorak or Colemak. It stayed around the same level. Which, again, is not dictated by the layout but by the accuracy with which one has trained their muscle memory.
Lastly, on the note of programming: It is true that for programming languages, on QWERTY the keys used often, such as {}, [], ', =, +, -, _, etc., are all to be reached with the right pinky, which drags performance down. This still is not worth the switch to something like Programmer Dvorak, however, since, especially in programming, the limiting factor is rarely typing speed (once you get above 60WPM, that is).
So given all this, there are also a few downsides to switching that I wish to elaborate:
Dvorak suffers from the huge disadvantage that all computers use shortcuts (such as the famous CTRL+C and CTRL+V) which, on the Dvorak, are in different and hard-to-reach positions. Colemak doesn't suffer from this as much, since it kept the C, V and B key positions from QWERTY. However, even with Colemak, using programs which rely heavily on shortcut use (the most notorious of these being software like vim and emacs), has to be relearned from scratch.
Switching takes a very long time. Let nobody fool you. If you were typing at >80WPM, I can tell you from personal experience that it takes months to achieve this speed again. Even if you swap only a few keys (like Colemak), it is still a painful and long process.
When you successfully switch, you will be unable to type fast on regular QWERTY keyboards anymore (take my word for it). You will still be faster than someone who doesn't use touch typing, but if you ever have to type on a QWERTY computer as an alternative typist, you will be in for some embarrassment. This can get especially hairy if it is work related.
Many alternative layouts are not nearly as standardized as QWERTY. In other words: If you use an older machine, for instance, you may find your preferred layout not installed. This is a further hassle, because then you have to get around that problem by downloading and installing the layout you chose, meanwhile having to work in a layout you can no longer use.
Thus, in conclusion, my advice is: If you are happy with your current layout, keep it. The benefits of changing are much too small to consider. Especially if you are a QWERTY typist, I recommend staying with it. It will save you a lot of hassle and annoyances.
Long story short: QWERTY was not designed for touch typing. When compared with nine other layouts, QWERTY comes in dead last in typing effort. Give other layouts a try and experience the difference.
Whether the QWERTY layout was intentionally designed to slow typists is irrelevant. QWERTY is inherently slow by design.
The ultimate argument made by alternative keyboard advocates is that QWERTY is an inefficient, sub-optimal keyboard layout and that better layouts exist that suit modern touch typing technique. Whether one subscribes to the belief that QWERTY is a contributor to carpal tunnel is another story, yet many anecdotal stories exist of people who have reduced RSI by switching.
You should use a non-QWERTY keyboard layout because a non-QWERTY keyboard layout is easier to use. Switching to Dvorak or Colemak reduces distance traveled, reduces same finger typing, increases hand alternation, increases hand balance, and optimizes performance according to finger strength. Switching to Dvorak alone reduces effort by 30% (i.e not 4% as suggested, see first link). Note that these benefits accrue over time with use. The idea that there is no benefit to switching is simply untrue. And with newer layouts such as Colemak, typists don’t have to relearn to type. 48% of the Colemak keyboard layout is identical to QWERTY. Even if one’s error rate doesn’t fall (highly unlikely), alternative keyboard layouts like Colemak remap backspace to an easier to reach spot.
Regarding said downsides:
Windows and Mac offer the ability to use familiar QWERTY shortcuts while pressing Ctrl/Cmd. Many people have devised ways for using emacs/vim while switching.
Switching can be very painless and easy considering there are websites such as keybr.com that teach how to master touch typing using Dvorak or Colemak.
Whether you can type as fast with QWERTY after switching is irrelevant. Once you switch, you are set and won't have to go back to QWERTY on your own device. Dvorak and Colemak are standards or can be easily installed on to all major operating platforms. If you are going to use others keyboards and aren’t able to switch the layouts, there are numerous online keyboard simulators.
The hassle of switching the layout on the computer is very minimal if nonexistent compared to the hassle of typing on QWERTY.
Advice: The benefits of switching are too large not to consider. You have nothing to lose by trying; if it doesn’t work out, you will always have QWERTY to fall back on. Let your keyboard work for you, not the other way around.
My two cents worth...
I switched from QWERTY to Colemak a few years back, mostly because I like to try out new things but also because I was getting a bit of RSI from all the typing I do.
I found the transition took a few months, with the first few weeks being most painful -- I took a lot of handwritten notes in meetings in this period! But perseverance paid off, I'm now typing at about 60-70 wpm, which is faster than my old QWERTY speed, and I can touch type properly, plus I don't get the RSI pains any more.
Now, much of this might be because I learned to type properly in Colemak (using programs such as GNU Typist) whereas I don't recall ever learning QWERTY properly. But I would say that Colemak definitely feels nicer on the fingers, with less movement around the keyboard for most words.
I'd also add that Colemak is supported as part of the base OS on Mac and popular GNU/Linuxs (e.g. Debian and Ubuntu), and easy to install on Windows, so would consider it fairly mainstream.
I agree with #mark-anderson I am typing on dvorak and the worst part is that shortcut keys are a pain - I use a tool for that (see my answer at https://stackoverflow.com/a/22945703/18132). Using someone else's keyboard is also a pain.
On the plus side, I can touch type and never have to look at the keyboard. actually if I look at the keyboard I get confused since the keys are in the wrong place.
Was it worth it? Not really sure. I like that I can touch type. I like that I can use qwerty shortcuts but my hand still hurts and I think I might be faster - not really sure. But I am considering switching back to qwerty - I have been dvorak for 2+ years now, so I don't really have a good reason to switch back other than to "conform" :)
I recently broke my finger and can now only type with my right hand. This has seriously impacted my typing speed. Since I write software for a living, this is a serious problem.
I have been doing some research, but haven't found a great solution yet. Here's what I've come up with:
Wacom tablet + hand writing recognition software. Is it possible to write code with hand writing recognition software?
one handed keyboards -- I have only found expensive (> $100) keyboards. These look like they have a steep learning curve.
one handed typing instructions: http://www.aboutonehandtyping.com/manualcompare.html. Does this really work?
What do the one handed coders out there use?
If you're a two-hand touch typist, the answer is a "mirrored" layout.
Mirroring lets you begin touch-typing with one hand almost immediately. Pretty crazy how easy it is. Based on the muscle memory you already have.
If you're typing with your right hand:
Type all right-hand keys normally.
Don't type left hand keys. Instead type the same motion (but mirrored) with your right hand.
So if you want to type:
"D" -> type "K" instead.
"W" -> type "O" instead.
"S" -> type "L" instead.
Same row of keyboard, same finger, same motion. Your muscle memory can already do this... kind of like how you're unable to pat your head and rub your belly at the same time. The wires in your brain are crossing somewhere.
Software to mirror the keyboard as described above:
Hold Spacebar to mirror:
Linux - MirrorBoard
Mac - Mirror-QWERTY
Windows - AutoHotkey version of Half-QWERTY Half-Keyboard
Predictive Text; Automatic Mirroring
Mac - One-Hand Keyboard
Windows - One-Hand Keyboard
Regarding one-handed keyboards, I've tried using a frogpad and found it ok for typing text, but unusable for coding. The symbols require several consecutive key presses and I found it impossible to use shortcuts reliably. It was too easy to hit the wrong key and get it stuck in the wrong mode.
Nobody has mentioned ENTI-key aka Coffee++ Layout yet? It is exactly designed for programming with one hand (left). And unlike qwerty, it is even optimized for speed and ergonomics. I used it some years ago for a short while and I don't know if it still works on newer systems. I think I used it for writing CSS: Typing the words with left, typing all those numbers on the numpad with right.
I can not recommend pen+handwriting. I usually use a tablet PC and handwriting code is terrible. I tried it on Windows 8 and Linux with Cellwriter, and both are not bad programs, but I still switch to onscreen keyboard whenever I can. But maybe the problem is my scratchy writing :)
I also can tell from experience that learning a new layout is not as complicated as it sounds. Especially if the layout is more logical than qwerty. I use Neo Layout since 10 years and getting the hang of it went smoothly, I was able to write a blog article after an evening of training.
"But what if you have to use qwerty on another PC?" This, also, is no problem, really. My simple trick is to never look at the keyboard when using Neo, but glimpse at it when qwerty-ing.
Good luck to anyone who wants to or has to use one hand for typing!
Now, the time to heal a broken finger will be shorter than it takes to adapt to one handed coding, not to mention the time it takes afterwards to get back to two-handed coding
Also, the time it takes to learn the methods is time you could've spend on coding (read: making a living).
Knowing this, we need a quick-fix, short term solution.
First of all, A good IDE, with code completion and similar functionality will help you a lot.
Secondly, use the shortcuts of the IDE, remember, there are Shift, Altand Ctrl keys on both sides of your keyboard.
(you might want to create a cheatsheet for those shortcuts)
In addition to helping you during your time with your injury, learning the shortcuts will also improve your coding speed when you're back up again.
Now, my comments on your proposals:
Don't, simply, Don't, it'll take even more time to fix writos (typos) beacause recognition will be flaky.
That learing curve will slow you down even more.
Won't even comment on that one...
Mirrorboard
A friend of mine broke his wrist snowboarding, and he had reasonable luck using speech recognition software (Nuance Dragon Naturally Speaking). It worked quite well for email and documentation, which would solve a part of your problem.
Another colleague, Nils Klarlund of AT&T, developed a version of emacs hooked into speech recognition. He even had a home-brewed set of foot pedals for doing shift, control, etc. He used this exclusively for years (due to bad carpal tunnel syndrome).
And maybe your feet can take up some of the burden. This is part of a parallel discussion going on in this question.
And off-topic, but extremely interesting, T.V. Raman, who's been blind since the age of 14, wrote a version of emacs that works with keyboard input and audio output. There's a chapter on it in Beautiful Code. I've seen him use it, and it's completely awesome. And of course emacs is a great interface for more than just text editing.
If you anticipate that your left hand will be out of commission for a long while, and if it's worthwhile for you to learn a new layout, then there exist one-handed Dvorak layouts.
There's some information at PC Guide: Single-Handed Dvorak Alphanumeric Layouts.
There also once was software for Qwerty Half Keyboards that used the space bar as an extra shift key that reversed the keyboard.
Good luck with your injury!
We have a developer in the office that lost mobility in his right hand and probably won't gain back full use of it. He has mainly learned to type well with his left hand and kind of fill in for his right hand. Although he lets his right hand kind of peck for things. He has gained enough speed back for it not to affect his day too greatly from what i can tell.
Only thing i can think of that might let you speed up some while typing with one hand and maybe being able to get a key or two with the other hand might be to use an IDE instead of text editor if you already don't, so you can use tab completion. Kind of a lame solution if you don't like IDEs or just don't have that option in your work environment but might help out a bit.
The same thing happened to me (I destroyed my left pinky). At the time, I didn't touch type, so my only use for my pinky was left-control, left-shift, and caps-lock.
This sounds as if it just happened to you. I promise you'll quickly learn to compensate. Remember, it's quality, not speed, that counts most.
Perhaps you should seize the opportunity and read to improve yourself as a programmer. Or spend some time debugging.
I consider myself somewhat familiar with Vim,
hate the arrow keys (let alone the mouse),
regularly look up tips and plugins to get the most out of this tool,
use it daily to manage my cloud servers, etc.
However, I always find myself doing the same mistakes probably inherited from the GUI-world:
too often switching to visual mode to see what piece of code I'm about to manipulate,
undoing changes to retrieve lost statements because I forget to utilize registers (and pasting code on temporary lines just to grab it after other edits),
relying on Ctrl-C & Ctrl-V when working with operating system's clipboard,
keep pressing j button to browse through lengthy files to find certain functions.
Probably my Hungarian keyboard layout prevents me from being faster as most of the special characters (/, [, etc.) are only available as a key combination (with Shift or Alt Gr).
Given this specific situation, what pieces of advice could you give me? Have you faced similar bad habits when you were a Vim-novice? I'd like to see my productivity skyrocket (who wouldn't?). Thanks in advance.
I've found a simple, effective strategy. Choose one action, one task or one set of keys that you think is unnecessarily slow. Figure out a better way of doing this using the vim manual or googling or a plugin etc. Force yourself to use this every time. Rinse, and repeat. The path to efficiency is one-by-one elimination of the slow parts.
I'd also recommend just reading the vim manual from time to time - even if you don't remember everything, knowing something's out there is very helpful.
This probably applies well beyond vim, but
something that worked for me was finding a specific feature that I knew would
be more efficient and concentrate on using that for a week or two.
Just one feature at a time, and possibly using it excessively.
After a couple of weeks it becomes automatic and you can move on to the
next thing.
I learn programming tricks the same way. eg. I'll have a month of using lambda expressions for everything, then a month of mapping and filtering.
(not on production code though)
Probably my Hungarian keyboard layout prevents me from being faster as most of the special > characters (/, [, etc.) are only available as a key combination (with Shift or Alt Gr).
I'm sitting in front of german keyboards all day long and know this problem very well. Some keyboard layouts are simply not very suited for programming / using vim. I think its safe to assume that most programming languages and keyboard shortcuts were designed with the us-layout in mind.
My advice: reset your keyboard layout to us-english and practive touch-typing on that layout (typing without looking at the keys). It won't matter that the keyboard labels are wrong and you'll be much more comfortable using vim hotkeys.
The only problem that still remains for me is to produce language specific characters (german umlauts such as ä,ö,ü) wich i assume will also be a problem for hungarian. For that I use a combination of vim-digraphs, linux window manager digraph-key and windows layout-switching hotkeys.
just keep using it. The more you use it, the better you become at it. VIM isn't too bad. The main thing is you just have to remember that it isn't always in edit mode.
Closed. This question is off-topic. It is not currently accepting answers.
Want to improve this question? Update the question so it's on-topic for Stack Overflow.
Closed 9 years ago.
Improve this question
In a discussion with co-workers today, I lamented that I can't ever remember what an icon means, and have to hover over them to see the tooltips, and thus to find the button I need.
On their side, they were saying that when the text needs to be translated, it might not fit (German vs English for example), and that every place where there is text, including tooltips, it needs a translation. So plain icons are easier.
What is the best tradeoff in useability for the extra work of text vs the subset of users who are icon-challanged?
I personally prefer text and hate icon-only UI's. I know that other people think the other way, equally strongly, either because of internationalization or because their brain works more rapidly with images than with text. If you choose one or the other exclusively for your UI, then part of your user base will be unhappy with your choice. (Sometimes this is the right choice, depending on how extensively the UI will be used.)
Internationalization is really not that difficult, except for finding a firm to do good translations of your text. The programmer portion of internationalization is pretty straightforward. However, I've known a number of programmers who prefer the all-icon method as it's less work. I've personally had to replace one all-icon-no-text UI that the users didn't like. The users said they could not remember what the icons meant.
I think more typically, many advanced users will prefer icons and many beginning users will prefer text. However, a number of advanced users prefer text. IMHO, any good UI will provide tooltips, so you need to translate your interface no matter what you do.
The most friendly solution is to offer both text and icons, possibly with a settings choice to disable one or the other.
I worked with people in a Human/Computer Interactions group and was raked over the coals for using icon-only. They had studies about comprehension, error rate, and speed of using UI's and a good icon/label combination won every time, all else being equal.
Localization should be a non-issue. You may have to localize the icons anyway and localizing a label (as long as it's stored as text and not as bits) is easy. In terms of size in the UI - that's another matter entirely. If you can't fit the text, I'd claim that your UI is too cluttered.
Really they both have advantages and draw backs. Text must always be translated into different langugues, and sometimes a single word will not be able to effectively describe the action of a button. For example, how would you describe the X button which closes a window in Windows. We know what it does, and most people I know call it the X button, but it doesn't describe what it does. It's a lot easier to put a button with an X symbol (or icon if you will) than put something like "close window".
That being said icons also have drawbacks. As you eluded they may not always be clear what they do. The user has to be able to put the icon in a social context to understand what it does. This may not always be possible. Also, icons in one language may not be understood in another, leading back to the translation problem. Icons can be advantageous in certain areas as they can take a complex concept and show the user with a small amount of space. (like to take a picture show a camera, or delete something showing trash can).
The trade off is really in a case by case basis. If you have power users who really understand the application they are using and the surrounding subject, you are probably ok with using them. If you have people who use computers once a month and don't really care to learn it is going to be confusing. Its the amount of information you can convey with a single symbol (icon, picture, letter) vs. the potential frustration of the users and the overall rejection of your program.
Make sure you have a way to get both. Screen readers have a horrible time with icons.
I hate icons, because I never know what they mean even if they're perfectly intuitive (like this world icon that means hyperlink above this box in which I'm typing). Several Unix terminal applications provide a choice between:
text
graphical icons
both
That's nice. I usually like the text on a prominent button, because the meaning of the button is much more clear and the mouse target can be a little bigger.
Its a cultural thing as some symbols (icons) mean different things to people of different cultures, backgrounds and experiences. There are global symbols that one could assume to be 'known' by the general population, i.e. the 'save' icon...
It is a fine line, but i think the tooltips are a good way to help out those who dont understand the meaning of the icon. Perhaps a set of options to have the buttons (icons) render with text instead of the icon image ? This could be a user preference in the application.
Perhaps a good reference would be some of the "extensively" used icons in microsoft applications such as 'Word'. I am generalising here, but microsoft applications are almost eveywhere and they have done all the R&D into suitable / effective icons.
You don't mention if this is for a web application, but if it is then you have to provide the text at least as a backup if the user has disabled images, is using a screen reader, or other limited interface.
two things i guess
the decision should be the result of usability research and properly quantified, rather than a dev's gut feel or whim.
an icon that doesn't carry an obvious meaning is a bad icon and should be changed.
all that said, IMO: Icons with a tooltip/mouseover text equivalent, with bonus that this can carry a reminder of the keyboard shortcut.
(Note: I use "button" here to mean "the UI element on which the icon and/or text is located.")
I think in almost all cases it's important to include text either on the button itself, or at least on a hover-over tooltip on the button, so that in the event that the icon's meaning isn't intuitive to a particular user, the user can find out the meaning by reading the text. (Note that the translation work still needs to be performed in either case.)
A typical case for not including text directly on the button itself is when space is at a premium; when you want to fit a lot of buttons into a small area. Examples include the "toolbars" used in many desktop applications, and also in some web applications -- for example, the buttons that appear just above the StackOverflow answer text entry field!
A good case for including icons is when the button doesn't always appear in the same place, and the user would benefit from being able to quickly visually scan for where the button is located. For example, if I have a lot of programs open on Windows and I want to quickly find my instance of Firefox in the Windows taskbar, I'll look for the little orange icon, rather than reading the text on each taskbar button.
I find the concept of the programmable, tiling, keyboard-focuessed window manager ion3 very appealing, but I think it takes some time to customize it to your needs until you can really evaluate this totally different UI-concept.
Therefore, I would like to read reviews of people who tried it for a longer time as environment for programming (in particular using emacs/gcc).
(The policies of the ion3-author concerning linux-distros are not easy to follow for me, but this should not be the point here...)
I use ion3 daily. It's a wonderful window manager. The tiling interface really enables you maximize real estate. Once you get it setup to your liking, it is much more efficient to navigate via the keyboard. Even moving applications between tiles isn't that hard once you get used to the tag/attach key sequence.
With ion3, Vimperator and the various shells I have open during the day -- I barely use the rodent.
The author's opinions aside -- a good resource for configuring/extending Ion to your liking can be found at:
Configuring and Extending Ion3 with Lua
I've been using Ion daily for nearly two years now. Good things:
Easy to use from the keyboard.
Handles multiple screens (Xinerama) very well (once you have the mod_xinerama plugin), especially as in my case the screens are different sizes.
Very predictable where windows will appear.
Splitting, resizing and moving windows is very easy.
Multiple, independent workspaces on each screen.
Very fast and reliable.
Bad things:
Too many different shortcuts. e.g. there are separate keys for moving to the next tab, next frame, next screen, and next workspace.
Applications that use lots of small windows together work really badly (e.g. the Gimp) because it maximises all of them on top of each other initially.
Sub-dialogs can cause trouble. Sometimes they open in a separate tab when you want them in the same tab, or sometimes the open in the same tab and take the focus when you want to continue interacting with the main window.
These things can probably be changed in the config files, but I haven't got around to it yet. Also, the actual C code is easy to read, and on the few occasions where I've wanted to fix something it has been very easy. I don't feel tempted to go back to a non-tiling WM, anyway.
I've used it off and on for the last few years, I think its a great window manager, but I keep crawling back to kde3 whatever I use.
Its however difficult to put into quantifiable terms why this happens, but its right up there with the gnome-vs-kde battle. Neither side can understand the other.
I would also just love to have kicker + ion3, but they don't gel awfully well.
Moving applications between tiles ( something I tend to do lots ) also is a bit inefficient ( too addicted to the mouse )
( Kicker + Evilwm is a good combination, but evilwm just can't handle stacking in a user-friendly way )