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)
Right from the start I think that the free software movement was geared by geeks toward geeks, and it's become a game of catch up developing quality gui's and moving away from traditional cli's. Added focus in this area will surely see a boost in gaining mainstream desktop users and subsequently enthusing them with the free software ideal. Of course a pretty interface is excess baggage and can lead to bloatware, but surely it's imperative to the future of open source?!
A lot of people do not understand how themes and options do decrease the usability of the UI. If you are able to improve the UI significantly by changing a few options or applying a skin, then there was something wrong with the whole thing in the first place. People should not waste their energy on themeing engines or messy options dialogs (my favourite horror is the KDE control center) but focus on one UI style and make that the best. Rather one perfect UI than a dozen so-so ones.
The attitude of free software nerds towards useability is often the defining limitation of the spread of free software.
The fact that it's free should not mean that you should have to be a nerd to use it. Good useability is probably more important than correct functionality.
To quote Bertrand Meyer from his monumental book Object Oriented Software Construction:
Good user interface designers follow a more prudent policy. They make as limited assumptions about their users as they can. When you design an interactive system, you may expect that users are members of the human race and that they can read, move a mouse, click a button, and type (slowly): not much more. If the software addresses a specialized area, you may perhaps assume that your users are familier with its basic concepts. But even that is risky.
Which lead to the following design principle:
User Interface Design principle
Do not pretend that you know the user; you don't
As obvious as these sound, here are my rules for a good user interface on the work I do, whether it's a web page or a GUI app:
1) Keep It Simple Stupid
2) Think Like a User (the dumber the user, the better the UI)
3) Make it LOOK better than it WORKS
The last rule is true because if the app looks like crap, no one will be happy, but if it looks GREAT and works so-so people (users, your boss) will be excited and give you more time to get the bugs worked out. This is my main problem with someone like Jakob Nielson's whiny biatching... the app has to be simple, but it's got to have good design too. Colors, buttons, subtle eye-candy, well balanced spaces, etc. Usability is key, but design is always first.
Anyways, many of the people who work on open source software seem like the people who CAN'T STAND USERS. And by users, I mean the stupid, stupid people who might use their software. It seems pretty weird, but it's true... Maybe a little less antipathy for the newbies would go a long way to helping OSS GUI design.
-Russ
Me
Though geeks like to tinker, there's something to be said about every application being consistent and keeping preferences, as much as possible, to a minimum and also to put them at the system level so they can easily be changed in a way that will modify all applications.
Its nice to be able to sit down to a Windows or MacOS system and be able to start working right away, not worrying what crazy menu is bound to your middle mouse button in application XYZ vs application ABC or where the person who configured this PC hid some other menu you need just because the preference system allows him/her to move it whereever.
I realize one size doesn't fit all and that geeks will always want to tinker, but I think if anyone is to take Linux/UNIX seriously on the desktop, there needs to be one clear Linux UI standard that is followed, and the rest must be relegated to 'alternatives'. This is similar to how Windows has its UI, but you can (if you really want) go swap it out with litestep or windowblinds or whatever.. The option is there, if you really really want it, but there's also a clear default UI system that companies and groups can standarize on.
However, the Man-Machine Interface (MMI=GUI+CLI) is the front-ent between me and the computer. Better yet, is _the_ computer as I see it. And what I like about Gnome is that it is a reasonable good platform with a lot of components (panels, menus, applets) with which I can build _my own_ GUI. Which is surely not the best GUI. But is it the GUI which best fits my needs.
I whish good luck to Gnome developers in their quest. I just hope they don't loose the component approach they had up to now. And don't lock users in _their_ idea iof the best GUI (while I agree that too many preferences are evil).
Ciao
----
FB
I 100% agree with this. The QWERTY example is a perfect analogy. We all need something similar to what we work with right now.
I say EMBRACE and EXTEND. Copy Window's UI down to the pixel, get everyone to use *nix and then when M$ changes their UI again, everyone will say "why doesn't Windows work like my Unix box?"
Think about the hardware equivalent of this: Compaq started creating IBM clones which was great, but eventually IBM tried to move to the PS/2 and Microchannel, yet everyone by that time was used to the open architecture and balked at IBMs proprietary play. The same could be done for software, but first you have to make a perfect copy.
-Russ
Me
This is one of the reasons why preferences are as common as dirt - you have to write code to make them work.
It is also the same reason why we have dozens of sendmail/ftp/irc etc... replacements out there. We don't need them, there are plenty already, but coders feel they have the need to code rather than feeling the need to think about what they code.
In the last few weeks, my hosting co's ftp software has been randomly giving me errors that suggest it doesn't know how to list a directory, put or get a file. Not that I need any of those features anyway, so I did some research and ended up installing ncftp (Mac OS X installer pkg). I realize ncftp's not a new program, but I am amazed.
It has everything I've ever wanted in an FTP client: speed, easy-to-use "bookmarks" (no more dumping passwords into clear .netrc files or entrusting them to Apple's security-hole-prone Keychain), status reports on transfers, and I can even use wildcards to up/download a whole mess of files at once without having to sift through ftp's man pages. Everything works intuitively, and I suspect there is much more I will discover just by using the tool.
I guess that's what a great UI is -- one that you can use and learn without having to RTFM.
(Before you reply in defense of the RTFM concept, I agree that there are types of software that should not be used until one has RTFM, but it doesn't hurt to give the FM a great UI.)
It is strange how many OSS fans relate to such things: On the one hand, they do not accept strict guidelines that tell you how an application has too look or behave - for them it is taking away their "freedom". OTOH, at the same time they're very strict and disciplined when it comes to following technical guidelines like RFCs, IEEE and W3C standards. No, I did not say hippocrats. But I did think it.
one of the common mistakes, is the thinking of a UI as the way your window-manager, desktop, and toolkit look. but one of the most common problems, like setting up hardware, and peripherals, basic system-configuration should be the place where linux-developers should get their ass up. and develop standards please, no one has thought, apart of gnu-step about how applications that users-install should be handled.
.
most importantly system-configuration has to be made much easier.
and someone should develop a GUI and interface for cd-burners, that has NOT to be run as root
as one could say its time we make nails with heads.
While I agree with much of what you write, I think the distinction between "average" users and "power" users is an important one. (I dislike "beginner" and "experienced" because that's not really the issue; often, some beginners will immediately want to learn all the tricks, while some experienced users will never need them.)
To take a prime example, and one of my pet peeves, consider the interface for your file system. On many systems, you can have a CLI with commands to do things like copying or renaming files, creating directories, etc., and you can have a GUI interface with folder windows and so on. Most users are happy using the folder windows most of the time, experienced or not. I'd much rather browse my file system with something like a tree view and file list combination than with a CLI and constant use of ls or dir commands, and I doubt I'm alone here.
OTOH, suppose you want to create a new folder at me/one/two/three/four/new, and currently all you've got is me. It's faster to open up the CLI and use a command that can make five directories in one go than to create one in the GUI, open it, create two under it, open that, and so on. Anyone who's tried to rename "*.doc" to "*.bak" using a windowing GUI is probably familiar with the problem, too: a few seconds at the CLI, a few minutes in the windowing system. And hey, when's the last time you saw regular expressions being used in a GUI? ;-)
So I think it's fair to say that some applications are universally required in some form, and that sometimes, different perspectives on the same functionality are more useful to different people (or to be more correct, are more under different circumstances, and it just happens that some people encounter those circumstances more than others). They key thing, though, is not so much the functionality -- you can rename a file in either a CLI or a GUI -- but the way(s) you provide that functionality to the user.
If you disagree, post your argument. (-1, Overrated) isn't your personal censorship tool for views you don't like.
Yes.
The goal should be the exact opposite: the have a system so intuitive and powerful that customising it is entirely unnecessary. Of course, you'll never reach that goal, but it should always be the target.
Your post was right on the mark.
If you disagree, post your argument. (-1, Overrated) isn't your personal censorship tool for views you don't like.
Good article. Take for example KDE, nice desktop but too many features. I posted before a comment to the KDE guys that to they should focus on deleting features and just choose good defaults. I am not using KDE anymore, went back to WindowMaker.
In any case, the general problem with OSS is that anyone can add patches/features, but no one dares to _delete_ useless ones. So in the end, a program just grows bloated with useless features. Delete is the key. I'd rather re-learn a few new well chosen key combinations, rather than having the option to make my 100 own ones.
The same argument also counts for the (Linux) distributions, they grow beyond proportions and expect the end-user to plough through the jungle of available features/programs.
The article lists these issues as why Free Software UI sucks.
Hello? Where is the #1 reason?
Sure, developers write things for themselves. Developers write things for their co-workers. But do developers of Free Software really go out and research the goals of their users?
In looking at Linux user interfaces, I see that most tools merely tie some toolkit strings onto the underlying code so that it can be manipulated. The current thinking seems to be that if the underlying driver can do something, expose that ability directly on the command line or in a preferences dialog box.
A great case-in-point is cd-burning software. Type (cdrecord --help). The typical GUI wrapper is just the Gtk equivalent of (cdrecord --help). A massive soup of options with little help for people who don't know what a leadin is, don't care what a TOC is, don't understand how the lovers Romeo and Joliet got into CD-burning, and don't understand whether they want to fixate the disc or not.
Instead, turn it over.
Make some archetypical example users of your application. Nate the newbie. Seth the secretary. Judy the junior admin. Devin the developer. Whoever it is that needs your help to accomplish their goals, get to know these people.
A useful CD burning tool doesn't need to expose everything the driver can do. Add music files here. Add data files or folders here. Might you want to add more files at a later date? Burn the disc.
In Alan Cooper's words, "don't make the user feel stupid."
A user interface needs to start with the user, and proceed to the interface.
[
Don't get me wrong, Apple do some good work with UI design, but to pretend that they have the most intuitive UIs is naive. When I first started using my friends Mac, he had to sit next to me and constantly remind me to close the app manually, which button means right click, and so on.
People waay overrate the Mac as GUI perfection. Open software shouldn't just copy it mindlessly.
Guidelines are guidelines. They are not hard and fast rules.
... need more examples?
/.ers about copying MS and how "wrong' it was.
Standards are standards. You screw with the implementation of a standard and your application is worthless.
You screw with the UI and it may be ugly or less user friendly, but it still works.
Most developers follow the basics regardless of OS - File, Edit, View, Help is always the last menu item on the menu bar. "File|Open", "File|Save", "File|Print"
IIRC, the last time someone emulated MS and followed guidelines (Ximian's evolution) there was an outcry from
Dammit, make up your minds!
And we won't even discuss the "Look and Feel" lawsuits in the early 90's when people emulated other UIs.
I don't have a solution, but I certainly admire the problem.
'Perfection is achieved not when there is nothing more to add,
but rather when there is nothing more to take away.'
(Antoine de Saint-Exupery)
You can increase the apparent simplicity and focus of an OS
simply by consolidating five menubars into one.
--| INTERFACE DESIGN > A SINGLE MENUBAR |-----
>> WHAT? Give me the summary.
A SINGLE MENUBAR AT THE TOP OF THE SCREEN that changes according to
the current context (window) instead of a menubar for every window.
Setting this as a User Default will improve Linux's ease-of-use.
Placing a single Menubar along the top of the screen:
1 - Makes it faster and easier to hit.
(no mouse overshoot to slow things down)
2 - Eliminates clutter in the interface.
3 - Reduces ambiguity (and hence - user error).
--| DISCUSSION |---
>> LINUX MENUS WORK GREAT NOW.
>> WHY SHOULD WE DO SO MUCH WORK TO CHANGE THE ACCEPTED DEFAULT?
In programming, if you compute a static variable within a loop - it is
highly innefficient - it slows down the loop. You optimize code by pulling
all the computes you can out of the loop and processing externally.
Interface design is the same. If a user has to click: A, B, C three hundred
times a day - it would make him 3 times as efficient to collapse those three
steps into a macro and execute with one keystroke. Making things less steps
for users optimizes the UI just like computing static variables outside the
loop optimizes code.
Since Menus are one of the most frequently used items in an operating system,
optimizing something small in this frequent behaviour equates to a Big savings
for the user over time. Therefore getting the menus right is one of the most
crucial and fundamental UI decisions that must be made by those implementing
the OS.
Linux currently imitates Windows' menubar implementation of putting a menubar
in every window. UI studies show this is not the optimal way of implementing
menus in an operating system. Linux can beat Windows in menubar GUI by providing
the option of a single context-sensitive menubar. There are several good reasons
for doing this:
1 - TARGETING CONSTRAINT
How easy it is to hit a target - virtual size.
2 - CONSISTENT PLACEMENT
How easy it is to remember "where" a target is.
3 - SIMPLICITY KEEPING FOCUS
Elimination of extraneous controls that are not
relevant to the current task at hand.
>> See the rest of this posting - Why Single Menubar
best regads,
john.
Hands together for the pat-your-own-back article for Gnome, thinly disguised as a discussion on User Interfaces.
We see this kind of crap all the time, seems that it's usually Gnome, but KDE tends to do it too.. A developer on (name your window manager here) posts a big ditribe about how (this or that) should be written, and behold, (our window manager here) does it perfectly!
Sheesh.
Remember, don't feed the trolls.
A perfect example of the problem with Linux for the average desktop user. Thank you.
FYI,
MOST people just want to get in a car and get from point A to point B. They don't want to deal with a manual transmission.
MOST people don't want to fly a 747. They want to get in one and get from point A to point B.
MOST people don't want to jump from a perfectly functioning aircraft.
As long as average people have to deal with that extraneous arcana to get simple (or even complex) things done, they will not use Linux on the desktop.
And when hardware and software support for desktop oriented technologies starts or continues to lag or fade because other platform providers do cater to the average user's needs, I will have no sympathy for you.
DCMonkey
The installer is always the first part of your UI that the user sees. Unfortunately, it may also be the last.
In many cases the question of "why should the end user be installing anything in the first place?" is totally ignored. No-one demmands that cars must be end user servicable. Or expects the typicall office worker to install their own network, telephone and power sockets, let alone assemble their own office using bricks, wood and plasterboard.
But suddenly when it comes to computers it's vital that everything be end user installable (even if it results in making the job of real sysadmins considerably harder.)