Making Linux Beautiful

intensity writes, "Wired has an interesting read on efforts to 'make Linux pretty.' I originally got into Linux because it gave me total control over what I could do with my desktop through the X Windowing system. That was in 1995, when setting up X was a chore in itself. Is it time to set down a standard GUI for Linux systems? " It's mostly yet another story about Eazel's Nautilis project (AKA the GNOME 2.0 Desktop)

File manager?

the article says about KDE and Gnome:
neither has a graphical file manager. Instead of clicking on icons or menus to open and save documents, users must type file names into a command-line interface.

But what about the KFM, KDE does doesn't it?
I didn't just dream about those pretty icons and stuff...

Standard?

Bruce Tognazzini, founder of Apple's famous Human Interface Group and a frequent interface critic, stressed the need for a standard Linux interface and a comprehensive set of interface guidelines to ensure consistency across the system.

Ugh, no. No standards on GUI, please. The great thing about using Linux is that there isn't just one way to do something. If we start forcing GUI standards on people, we'll get bloated window managers that don't serve the needs of particular people. It's very very difficult to make one interface that suits everyone without making it be this huge Swiss Army knife of a program (see emacs, although please, no flames).

The X standard is enough. I think Eazel's efforts are great, but I like the variety that the current crop of window managers provides and if you 'standardize' an interface or the commands it can use, inevitably, something will be included that you don't need and something else will be left out that you do.

Poorly researched

Although Linux already has a pair of evolving GUIs -- KDE and Gnome -- neither has a graphical file manager. Instead of clicking on icons or menus to open and save documents, users must type file names into a command-line interface.

Err ... yeah ...

Not just sittin' pretty

I think it's a mistake to think of a GUI as being only useful for "prettying up the OS".

Really, there are a lot of steps in certain routine configurations and obscure syntaxes that I, as a sys admin, would prefer not to have to edit by hand if I didn't have to. For example, I probably add/remove ten DNS records in a given week -- I'd rather have a GUI to do it. I doubt, however, that my boss would want to pay me to develop one and I'll be damned if I'm working on my off hours.

The mistake that most of the command line crowd (myself included) make is viewing the GUI v. command line debate as being an either/or situation. Well, just because that's how MS does it, it doesn't mean that we have to make the same blinding errors.

In any event, any Linux GUI will always beat the hell out on the NT version for one simple reason: setenv DISPLAY tux:0 -- while the rest of my officemates are driving upwards an hour to get to their clients to reset some little thing or the other, I can do anything remotely that I can do in person.

----

The Font DeUglification HOWTO

As much as I like Gnome, it is possible to make X a lot nicer looking without signing over your first born to any particular desktop environment... the nice thing about this is it's not "theme beautiful" - just "easy on the eyes" beautiful. :)

The XFree86 Font Deuglification Mini HOWTO [linuxdoc.org]
----

Re:Standard?

Your missing the point.

Linux needs a standard desktop of the same level as either Windows or the Mac for those who don't know what they are doing.

But, because this is Linux, the user can choose to entirely replace the desktop with what they choose (unlike the Mac or Windows where you have no choice).

So this is a win for both sides. The people who just want to use a computer with a familiar interface get their interface, and the power users get the benefits that increased market penetration bring.

agree, clueless

As most other posters (so far) seem to be saying "These people are cluseless about unix."

Okay, there were no graphcisla file mamagers in 1982, but come on people this is 2000. Sun has had graphcics filemanager since they were trying to get everyone to sue their own propritary window system over X. They have had one for X for years. That is over course one example.

I happen to like unix, with the many different approaches that work togather. If they were working on a standard way to make all file managers interoperate (so I can drag a file from KFM to the irix dumpster for example) more power to them. But to say that I should use their file manager over the 100+ out there? Forget it, I'll use a different one just to be contary.

I don't think these people understand that unix still has a following amoung people who don't care if anyone else can use it or not, only that they can get their work done with speed.

To put it anouther way, my prefered default shell is csh Not tcsh, sh, ksh, zsh, bash, or any of a number of less well known choices that all get the job done. Yeah you can make arguements against csh, but I'm used to those idiocryncies and there is no clear advantage of switching. (of course when I'm root I use sh) Likewise I use tcl not perl or python (either of which is better suited) to write my simple scripts. (again, scripts for starting up the system are sh - though now perl is a part of most base systems)

How mass market does Linux really want to be?

"But of all the things in Linux, this is the thing that's missing. It must have a standard interface if they want the revolution to go all the way to consumers."

I by no means have the answer to this, but at the very least I'd like to pose some questions to the crowd: Is our goal as linux users and open source advocates really to take Linux et. al "all the way to consumers"?

My feelings are that we use and love Linux for various reasons - the community, the flexibility, the attitude, the speed, etc. Linux is a more powerful and more stable operating system than Windoze. X helps us visualize that operating system and GNOME/KDE help us build on that visualization.

But frankly, most "mass-market" people don't know, care, or want to know how to use *nix. It's complicated, and that's why, as hackers, enjoy it. The problem arises in that there is an inherent loss of control/power (i.e. Windows/Mac) when a system is "dumbed down" beyond the window manager.

So what do we want? We want acceptance of Linux as an alternative to the norm. We want wider application support. We want games. But I don't necessarily feel that taking Linux "mass-market" is the only way to get those. (And I do agree with the only ways Linux could get really "mass-market" are a) make it really dumbed down or b) make everyone in the world suddenly get a lot more tech saavy - unlikely).

I'd be curious to hear what you all think about this....

I'm beginning to REALLY hate GNOME

I don't run GNOME on my system because I like the feeling of "raw Linux" and because I don't like how GNOME looks/works/acts/etc. That's entirely up to me and isn't the reason I hate it.

What I hate is all the unnatural dependencies on GNOME on RedHat systems. For instance, last night I upgraded my RH5.2 system to RH6.1. I have a lot of complaints about how this worked (like, why can't I cancel or at least unmount my drive? and why can't password-less users login or at least have root change their passwords), but the relevant complaint is the GNOME deps.

After the install I found that a lot of GNOME stuff had slipped through (another complaint: when I DON'T want to auto-install deps, let me UN-install upward deps). I spent a few minutes rpm -e'ing these, but when I tried to remove gnome-libs it told me that wmconfig needed it. OK, so get rid of wmconfig. Can't, fvwm needs it. WTF? That's just not possible. fvwm is ~8 years old, GNOME is
Actually, the real blame for this goes to RedHat (for stupid dependency defs) and the RPM format (for not allowing "wishlist" vs "gotta-have" deps). So once I get a tape drive I'm going to back my system up and install Debian. I hope it's a little saner and less "user-friendly".

To bring this post marginally on-topic: I don't mind if some (even most) people prefer "pretty", "standardized" Linux as long as two conditions are met:

1) I can still get to the "raw" level that I like.
2) Apps are not written to depend on a "standard" and break when it isn't true. There's a lot of tools on freshmeat that I'd like to use, but they all start with "g" or "K"--so I can't use them.
--
Here is the result of your Slashdot Purity Test.

Antialiased fonts

The biggest advancement Linux/UNIX GUIs need is Antialiased fonts. Apple did it, now its our turn. X has a lot of great features, a lot of nice and easy windows/desktop managers, but text sure does look like shit unless you have a 100+dpi moniter. Even then it isn't very impressive. My girlfriend always asks "why do the letters look like that when they are small?" and I respond, "Because that is the only thing Windows has better than Linux."

Re:Standard?

Ugh, no. No standards on GUI, please. The great thing about using Linux is that there isn't just one way to do something. If we start forcing GUI standards on people, we'll get bloated window managers that don't serve the needs of particular people.

The worst thing about Linux is that there are too many ways to do something, and everybody and his aunt feels free to reinvent the wheel in his own way.

Well, that's not really the _worst_ thing, but it sure holds up people that wish to port some GUI-based application to Linux... if there's no commonly accepted API to call, you either have to include the whole GUI in your app - making it bloated and crash-prone - or you have to tie up to Gnome or KDE or whatever and pray that it either becomes the standard in the future, or at least survives long enough for your purposes.

Notice I'm not talking about visual appearances, but at least standard ways to set up menus, windows, and drag&drop...

Standards and interfaces

There are two kinds of standards: official and consensual (or de facto). To create an official standard GUI for Linux would be near-impossible, because nobody has the authority to dictate what standards must be followed. If Linus doesn't like something, it may not make it into the official kernel, but nobody can be stopped from producing their own GUI.

To create a de facto standard GUI, all you need to do is create a GUI that everyone wants to use. So, standardizing the Linux GUI in this way is easy: just create a GUI that is all things to all people, and *poof* it becomes the de facto standard.

If you think that it's not possible to create a perfect GUI that is all things to all people, then you must abandon all hope of creating a standardized interface. Fortunately, the whole notion is silly anyway; there is absolutely no reason that a myriad of interfaces cannot exist. Just because many exist does not mean that a business cannot internally standardize on one, or that every Linux user must know how to use every single one!

Re:I saw a ploughman

But none of it will make any difference, because X-windows is already dead. It died a while back when someone thought "gee, I know, let's make a desktop whose strengths come not from the elegance of the theoretical design of the comittee that programmed it, but from its usefulness to the world at large, and the average human being in particular."

The fatal flaw in your argument is that X, in itself, is the user interface. Of course you are quite wrong. The window manager is, and this can be changed and made to act however you want, even like MacOS X if you wish... Just go get enlightenment and the Aqua themes for it and GTK, and you're off and running...

X isn't dead, the popularity of Linux is actualling bringing MORE people into X. X has got to be the best windowing system just because you need a window manager, all other window systems are pretty much static, and you cant change how they behave or look.

So, from where I'm standing, MacOS X, if it ever gets released for Intel, has got it in the bag.

You can stop dreaming, it'll never happen.

-- iCEBaLM

Re:Standard? = tradeoffs, good and bad

Linux will not compete with Windows unless there is a standard interface. However:

• Linux shouldn't necessarily compete with Windows. And:
• A standard interface doesn't mean you won't be able to switch to a non-standard interface.

It's kind of like the RedHat distribution. A lot of people mistakenly think RedHat = Linux. While the notion is wrong, if people make their products work with RedHat, there's a good chance you can make it work on any Linux system, and still have all the configurability and customization you'd like.

What I think is that there should be a standard FRAMEWORK. Something where you can write a program to be compatible with the "desktop environment", and not necessarily Gnome or KDE or something else. I would have it be like OpenGL - vendors (in this case GNU or Troll Tech) could add extentions, which might eventually be rolled into the specification for the "desktop environment" if they are widely accepted.

The only problem, as with many open software projects (and can be a drawback) is who ultimately decides what features are required to be compliant with "desktop environment 1.0"
----------

Re:Standard?

Valid points, but wouldn't it be nicer if all GUI apps played fair and co-operated. I mean *all* apps, not just KDE with KDE and Gnome with Gnome. Just simple things like using the same keybindings for open, save, etc would make life less troublesome. It doesn't matter much if you use QT or GTK based programs then, since they can be made to look similar (with QT's new theme stuff). Its the interfacing that needs to change, not the toolkit.

A common interface is a must for the average Joe !

There are two trends fighting here : those who think nothing has to be elected as a standard interface , those who think it has to.

It's quite simple, though : if the goal is to make Linux only accessible to the happy few who know how to use a computer, and how this computer works internaly, choice is a good thing, no wonder about that.

But if the target audience is (as it tends to be these days) the non-computers-professionnal, the lawyer, the doctor, the student, with their precise needs, a common interface is a must !

I used to work in a netcafé during my holidays. The OS used there was Win9x. One day, a member of the staff decided to move the icon for IE some pixels at the right. A lots of customers complained because it was "impossible to surf the net anymore !".

Most non-computer-litterate users are lost when you slighlty move an icon, what happens if there's a different desktop on every computer ?

Then again, that only applies for those who want Linux to become a popular, widespread OS for the average Joe.

Stéphane

Extremist

Just for the records I would like to say that there are a couple of mistakes (lack of information ?) in the article, the first and weirdest one being that there are no graphical file managers for Linux. At the moment I am typing I am using Kruiser (and I have been using it for a couple of months now) which looks a lot like the Explorer of NT4 (only that it is better).
So don't tell me that there aren't any file managers
On the other hand (and here I speak for myself) I wouldn't even consider not having the CLI since sometimes it is more useful that any graphical tool (mainly because I can type faster than I can move the mouse - and I am really good at that too).
So, in my opinion, and I won't be the first to say this, a system that can be acccessed only via GUI is deemed to be a Windows. No matter of what is under the hood. And I'll give you an example. Say you have a huge list of files that you want to rename (e.g. from Prefix-*******.something to ******.something). I'd like to see the average user doing this by hand (especially if you have 1000 files or more). Of course, any Unix/Linux user could tell me to use awk (at least I would) and do the job in a few seconds. Where is the use of the GUI here ?
The second part (and I like to dream, only that this article was too much) is about a system that plays the secretary. Come on, guys, I've been working with computers since '90 and I can tell you that there is no machine stupider that this one. And believe me, as long as we don't find something really revolutionary (i.e. understand the human brain and imitate it in AI) it won't be possible to have this sort of GUI. I mean, look at the search engines. I happens so often to return weird results even when your query is quite well constructed. And of course, what you need is there, in the result, only that it is in the 2nd or even 3rd page. So, no matter how much I would like an intelligent interface, I think that allowing the computer to filter the information and hide some of it is very risky. It is very likely that an important message will be purged without giving you the chance to do anything about it. And this is only an example
Don't take my considerations as pesimistic - I simply don't think we can do this yet.
And, of course : the most important part of all. No, there shouldn't be only one interface. We are saying that in the bussiness a monopoly is harmful and leads to stagnation. Well, don't you think this is true in almost every situation ? Including GUIs ?
Or the author of the article is under the influence of the Apple strategy : we are the owners, we produce a unitary product, we offer only one choice. Period. Oh, I forgot : you are not allowed to modify anything.
How does this fit in the GPL/GNU/Linux/... scheme ?

This thread is scaring me

Many of the responses to this article tend to go like this:

"Who needs a fancy schmancy interface? Raw Xwindows is great!"
"We don't want to make Linux easier to use!"

Look, Linux started out as a private hack that went public and has grown into a very stable operating system kernel. But the long line of open source development and bleeding edge kernels is not the means, not an end. The goal is not for Linux to be the OS of the 3/_33t who like recompiling kernels and hacking X windows config files. If you want to be able to graduate and not have to use Windows on the job, then Linux has to progress beyond the embryonic stage.

Quite frankly, we need the expertise of people like those at Eazel. I recently tried out OS X for the Mac (which runs on a Mach kernel), and it's far from perfect but it's so far above and beyond anything for X, including MacOS-like themes, that you can't make valid comparisons.

Vector

You know, I think Apple has hit the nail directly on the head with their PDF engine for MacOSX. Vector based interfaces will be the future. Even if that vector art is rendered all purty-like to the point that it looks like raster art, I still believe that a vector based interface will prevail.

Why? I have a 19" Monitor at home that I use on my production machine at 1600x1200x32Bit. One of my roommates on the other hand, has a 14" monitor that he runs at 800x600x16bit. Look at Flash on the web. If I look at a properly done flash site, I can view it at full resolution, stretched to fill my screen, while he can do the same, and they look pretty much identical -- with the same piece of flash 'code'. I don't get pixel-chunky curves when the plugin 'stretches' the flash to full screen for me, and he doesn't lose too much quality when he views it at 1/4 my size, thanks to anti-aliasing.

As monitors get bigger and bigger, faster and faster, and we still need to (at least somewhat) support legacy displays, Vector based interfaces will play a large part in how we view our 'desktop'.

Re:Not just sittin' pretty

The administration issue is still a gaping black hole in the Linux usability picture. Yes, you can make your desktop "beautiful" with themes and whatnot; Yes, the installer is easier than Windows 98's. But, no, nobody has really tackled the administration issue -- things like linuxconf and the other tools you find seem to be shaky or ill supported at best.

The root of the situation seems to be ultimately cultural. Most people use Linux because they want a unix system, and unix systems have a long standing tradition of obscure administration practices. Folks have made your suggestion this board that you can have your cake and eat it too, that admin functionality can be integrated into the pretty GUI without minimizing or removing the unix environment. These suggestions are usually met with disdain -- the community as a whole seems either happy with the way things are now, or afraid of 'windozification'.

For some folks like me, it's part of the fun. But it does bring up the question of whether World Domination is really the goal or not. Widespread deployment of Linux, even on the server level, is going to require the enlistment and education of the current minimally skilled small server admin crowd. It's a tough problem -- the average MS certified goon isn't even qualified to run a mid-sized NT installation (I know, I've interviewed these guys, and apparently MS will certify you even if you don't understand basics like WINS or domain controllers), and now these guys are being asked to wade into quagmire of a unix installation without any tradition or dummies books to guide them.

Small and medium sized businesses are seriously considering Linux as an alternative to Windows, but even at zero cost, it can't do them a lot of good if they can't find/afford anyone to run the system. The barbarians are the gates -- the question is are they going to find a soft landing when they get in, or is the potential configuration nightmare of a unix system going to kill the prospect of widespread Linux deployment? Is World Domination really what people want, or is Linux best suited to the traditional unix niches of academics, ISPs, and glass houses? It's a tough problem, there's no clear answer, and it's something that may well lead to a fork in the community.
--

Example

Here is the letter I sent:

You have probably received 400 emails about this already, but just in case - here goes. Both Gnome and KDE already have graphical file managers, contrary to the article which said neither has a graphical file manager. Here [gnome.org] is a screenshot of the Gnome file manager and here [kde.org] is one of KDE's.

Please issue a retraction so people are not misled. It is very important to the Linux community that people in the more general computing public become aware that Linux is becoming easier to use. Thank you for your attention to this matter.

Re:File manager?

Or the Gnome-MC (which is, IMO, very nice). This line struck me as the most self-contradictory:
"users must type file names into a command-line interface."

Is it the common dialog of GTK+/QT he doesn't like? I don't recall the Win32 common dialog having a nice little "click pretty pictures to specify a textual name" thing.

I think what we have is a red herring article designed to trump up the new Eazel effort.
Compare:
"Nautilus' file system includes a point-and-click file manager that will offer advanced features like the ability to tell the difference between sound files and pictures, and display them accordingly."

This right after mentioning that the ex-apple people would be using Nautilus. It's practically a press-release.
Step 1: Say there are no graphical feature F in X and Y graphical front ends.
Step 2: Mention that Z project is using program B
Step 3: Crow about program B. Mention it has features F, K, and G.
Step 4: Mention that some guy related to Y front ends and other internals has "seen the prototype, and loved it" -- something which should not be special with opensource (ie: everyone can see it).

He also ignores the fact that Gnome-MC and KFM both use mime types to display file types (although KFM, by default, seems to know fewer mime types) claiming that they don't know what they are (by saying that Nautilus does).

Then the semi-FUD makes its appearance:
"questioned the need for yet another PC interface -- especially one based on a design that hasn't changed significantly in 20 years."

We don't rethink addition or subtraction every 20 years just because the knowledge is "old" and "hasn't changed significantly" .. I really wish people today would understand the concept of matured software/knowledge.

Then the guy is quoted as saying that they want "internet enabled" desktops which tell you what you need to know. Which strikes me as wrong because 1) if I want to browse, I open a browser and 2) if I want to filter knowledge, I can.

I use Linux/Gnome and Win95 on my workstations at home (Win95 for games and some apps, Gnome for the rest). And I've worked on machines with the "IE memory/CPU tax" installed at work. Considering the 3 to 5 second pauses most tasks seem to require as parts of IE load for doing things like file management on a machine that is 400Mhz (Celeron) and has 128mb of ram (SDRAM DIMM).. It's funny to see how many Win98 machines perform worse than my own slower machine at home (a paltry 266Mhz K6-2 w/ 128mb of ram SDRAM DIMM). I really wish people would get off this integration kick that MS seems to have started (and which Gnome, KDE, etc seem to be following a bit). The basic shell should provide mechanism, not policy. If I want to do a set of tasks, I'll have applications for them (that are small and fast).. Not some huge integration crud..
---

Lets talk aesthetics

I've been working on a Windows based application, and I was determined to make it as visually uncomplicated as possible, while following the general Windows/MDI scheme. While I think I succeeded reasonably well, I think I failed in an important respect: the interface is not very attractive.

Many people might not think it matters, so long as it is functional, but I think it does matter. The boss has always been able to have nice offices and furniture, and a generally pleasant environment, whereas the drones have had to make do with ugly but functional gunmetal gray desks and gray fabric cubicles. With software, there's no reason for this dichotomy. Everyone should be able to enjoy an aesthetically pleasing software environment.

Windows does present a bind; part of doing a good job on an application is that you don't really want to stand out in a garish way from the organization. A fairy tale castle may look beautiful on the Rhein but would be silly in Manhattan. Practically anything looks silly when plopped in the middle of Stalinist era apartment blocks. This may be as great a reason to detest Windows as any: it's unncessary aesthetic mediocrity. Microsoft has an incredible tin ear when it comes to visual appeal (have you ever looked through the PowerPoint themes?).

Visual appeal is not the only, or even the most important element in GUI aesthetics, but it is relevant. I tried the Aqua theme on my Windows box, and it made work just a little bit nicer. Unfortunately, the skins program had compatibility problems so I had to delete it. This kind of themeability should be a standard part of every GUI.

Re:Standard?

>The worst thing about Linux is that there are too many ways to do something, and everybody and his aunt feels free to reinvent the wheel in his own way.

The best thing about Linux is that everyone has the freedom to do things they way they see fit with whatever tools and methods they can find/usr/build... and everyone gets a chance to reinvent whatever they want, whether it be the wheel (though my box is on plastic feet), graphics APIs, or networking filesystems...

This is a Good Thing(TM) and while there should be consistent calls to the display subsystem (this is fairly standard) the toolkits, widgets and other stuff that you link in should be very flexible. As for inter-WM communication, there are moves toward standards there, and really, most KDE apps run in Gnome, and vice-versa (aside from the desktop management tools, which no longer apply anyway)...

It's a good, free world, take advantage of it.

Re:Desktop model is dying.

I never understood this "3-D UI" of the future. Our language is two dimensional, our books are two dimensional, our desks are two dimensional planes upon which we work. The best we could hope for is a 3-D wrapper for our 2-D world.

Audio can be recorded in 3-D, but it certainly doesn't enhance the ability of a speaker to communicate, it just sounds prettier. People today readily discard quality audio for efficient recordings. By the same token, why would people want to work in a 3-D rendered UI?

Personally I think the next useful UI will go in the opposite direction. Less virtual, more real. Desks which automatically digitize handwriting, panels which are as easy on the eyes and as portable as paper, and of course, lots of specialized devices, like palm organizers, crosspads, networked stereo components (ala MP3 players). We're seeing the beginning of this now. Transmeta, embedded linux, and the like are a step into the future.

Quiet, unobtrusive and asthetically pleasing. If you want to dictate, dictate, if you want to type, type. Conduct online research from your sofa, or at the coffee shop down the street. Offices will become passe as the mechanisms of communication become easier and easier.

On the other hand, if you want to play games... go ahead and imerse yourself in your thick 3-D virtual reality.

Re:This thread is scaring me

"The goal is not for Linux to be the OS of the 3/_33t
who like recompiling kernels and hacking X windows config files. If you want to be able to graduate and not have to use Windows on the
job, then Linux has to progress beyond the embryonic stage.

Goal? There's a goal now?

Seriously, though: I think you'll find there as many different goals as there are developers. Personally, I'm 100% happy with the UI on my (Linux) desktop today. The Gnome peeps are nice enough to want a Linux desktop for the unwashed, and that's nice too.
--

Send all flames to /dev/null

I tend to disagree with what everyone seems to be saying here. Just hold your horses, and breathe through your nose for a minute while I share my humble opinion...

I think there should be a standardizing of UI's for Linux. Everytime I bring this up everybody bitches and moans about how if we standardize the Linux UI it'll get "...as fuck ugly as Windows..." Let's get one thing straight - I DONT WANT TO STANDARDIZE THE GUI - Just parts of the UI. If we standardize the GUI, we'll lose our artistic expression. Personally I think the KDE kids are doing a nice job, but it's sooo damn ugy. That's why I use E(yecandy)nlightenment.

The reasoning for standardizing the UI is to remedy all the blasted different ways programs behave. For example, in some programs, to quit the hotkey is 'CRTL-Q', while in others its 'ALT-Q', or ':q!', or even 'CRTL-X CRTL-C'. And to save it's the same problem - 'CRTL-S', 'ALT-S', ':w' or that awful 'CRTL-X CRTL-S'.

Why can't we just set aside our differences and play nice?

Oh Puleeeze!

I'm sorry, but this anarchy-for-anarchy's sake stuff gets to me when you talk about Linux' UI (or lack thereof). I really do not intend that comment as a flame... ;-)

As a former Unix/C programmer who has backslidden into Management, I have listened to a LOT of folks who might like to use Linux, but just find the UI baffling. Believe it or not, these same folks can actually use MS windoze. Germaine to the article at hand, though, my users RARELY ever open up the old Windows NT Explorer (formerly File Manager).

ANYWAY, a set of UI guidelines (and that's all they're REALLY proposing, by the way) are long overdue. This, to me, is the ONE THING that now differentiates the Mac (and my old NeXT that I still miss, may she Rest In Peace) from PC's. Sit at a PC or Linux box and try to find something on the various menus. You just have to keep looking until you find it, look in Help, or (like my users) pull out your '45 and shoot the computer out of frustration.

On the other hand, things on a Mac or my dead NeXT are pretty much in the same place from program to program. This is really due to a document more or less called the User Interface Guidelines, IMHO.

Of course, QuickTime doesn't count. ;-)

Please Let's DO encourage this! I think they CAN be made generic enough to please most folks.

Standards are coming, voluntary or not.

I think it might be worth our while to admit that, as Linux is extended to a mass audience and an increasing number of commercial apps begin to appear, a standard is more less inevitable. The real question which faces us is this: Do we want to design that standard, or do we want it to be a hack? Worse yet, do we want a corporate mandate?

Stay with me here: Commercial app makers know that you can't please everyone all the time, and that people like to be able to fall back on things they've already learned -- Ctrl-S is save, right?

It is therefore in the interest of these commercial app makers to go with a the same system everyone else is. If a particular GUI has a larger marketshare and more consumer-ready apps already existing, that's where your company should go, right? Folks, we've learned this lesson before -- or have we already forgotten the one big lesson MS had to teach us last decade?

There will be a standard. Hell, MS could probably set those standards tomorrow by introducing parts of Office for Linux and saying "this is how it is". It's all about the apps. The question really facing we in the OSS community is whether we want to remain involved in guiding Linux, or if we want to give it over to the same type of people who've stalled the current level of innovation for the sake of a few (billion) dollars?

----

Re:Standard? = tradeoffs, good and bad

Linux shouldn't necessarily compete with Windows.

A lot of people, including me, will agree with you on this. I think Linux (and friends) got where they are today due to their disregard of Windows. The idea has not been to create a competitor for Windows. It was to create a stable, flexible, open, modular operating environment. The fact that Windows happens to be none of these things is what makes Linux a threat to Microsoft.

However, now that many people have become used to using Linux, they get disgusted when they are required to use Windows. I work on Windows NT boxes all day long and think how much productivity is wasted fixing things that never would have broken in the first place if we were using Linux or BSD. What bothers me most is that I have to click 20 different things to do something in Windows that could be done in Linux by typing a short, if perhaps cryptic, command.

That said, I'll really be happy to see better GUI utilities for Linux. Gnome's file manager (yes there is one, despite what the article claims) just doesn't cut it for me. I hate to say it but I'd rather use NT Explorer (not to be confused with Internet Explorer) than gfm. Fortunately I find it pretty easy to use the shell instead and type "cp file1 file2" to copy files, "mv file1 file2" to move files, and "rm file" to remove a file. So I rarely use a file manager anyway. There are times however when I prefer to see a visual representation of the file system...especially in cases where the filesystem is complex, or just plain messy. I understand why Nautilus seems to be Eazel's primary focus at this point.

BTW, in case anyone hasn't mentioned this, non-Linux/*nix users can get a look at many of the currently available configurations/themes for Linux at http://e.themes.org [themes.org]. I'm really just pointing this out because I object to the idea that people's Linux desktops are all ugly. True, the default Redhat 6.x desktops are as bland as Win2k's default, but there is a lot of flexibility already in how you can make it look. As a default setting, bland is better anyway.

numb

To hell with this!

I just want more buttons on my mouse :-)
That's the best UI improvement I can think of. One for open/save/compile/run/email etc.

Seriously though. Sure, have a default one. Make sure it can and will be customized like hell by the user.

Make the wm learn - MS Offices idea of hiding rarely used menu items was a (lightly misguided) in the right direction.
I want the wm to notice that I usually minimize XMMS within a few seconds of starting it playing. When it notices this it will ask me "In future, do you want the Play button to automagically minimize too"

NB: I meant "proper" data-mining. I don't mean the XMMS developers explicitly put this in. I want it to genuinely apply statistics to my habits. Notice the way after you save a file you quite often do a compile/execute. If it notices that this usually happened to files being saved in a directory called ~/c_progs (like I have).

I don't think it would be _too_ hard to implement. I have thought about this kinda thing in depth - well, knee deep anyway, not neck deep ;-)

There's no end to what it could learn. File dialogs could notice that certain directories are visited a lot and make them quicker to reach. It could not what kinda file extensions are associated with particulars dirs. If implemented properly it really could learn/customize stuff that the developers never expected!

Just maintain a big database of every "common" action the user takes. Events that happen within say five seconds of each will be stored in a list. Every so often (middle of some night) pump it into a program that looks for non-randomness. Companies like IBM are into this kinda stuff, AFAIR. Insurance companies would love it.

It's more a mathemical/statistical problem than a programming/coding one.

Blah Blah Blah. I must stop ranting (and going off topic)
------------------------------------------ -------
"If I can shoot rabbits then I can shoot fascists" -

a design that hasn't changed in 20 years

The article features that old anti-UNIX argument again: that it's design hasn't changed for so many years (and therefor is outdated - that's what these people usually want to suggest).

The Windows design is still changing - in a desperate attempt to bring the virtues that UNIX has featured for a long time to Windows.

The MacOS design is currently being completely revised to be based on - UNIX!!!!

IMHO, the reason why UNIX isn't changing much anymore is that it does all the things it's supposed to do very well.
Just like sharks - biologists say they've hardly evolved for millions of years now - because they fill their biological niche perfectly, there's nothing to be improved anymore!

What was that clever sig again?
"Those who don't understand UNIX are condemned to reinvent it - poorly!"
(or something along these lines

It's the metaphor, stupid

We don't need more work on making our OS pretty. (KDE for example is "prettier" out of the box than all the GUI enhancements for Windows put together.) And it's not that we need a standard GUI.

It's that the GUI, any GUI, has to have a standard metaphor that works across all programs. We have long since diverged from a standard metaphor from the first days of the Mac desktop. The metaphor was supposed to be pieces of paper that you push around on a desk... and which also have buttons and knobs and levers and sliders on them?? Is it any wonder your mother is confused? The mouse, which is supposed to be a metaphor for your hand, alternately becomes a metaphor for a pencil (word processor) a finger (the frame of the "window" - another blown metaphor), a volume control knob (don't even get me started on the windows 2000 cd player), and half a dozen other things.

GUI designers no longer understand the simple concept: one metaphor = no confusion. Human beings are designed to recognize patterns, but there's no single pattern in the modern GUI. If you make your GUI desktop a million different metaphors, it doesn't matter how pretty it is or how much command-line complexity it hides - if the human being who's never seen it before can't quickly extract the pattern, the metaphor, from it - you blew it. Compared to a "Windowing" GUI, the command line is easy to understand because it's like speaking orders to a butler. Do this, do that. It's a single, unified metaphor. It's hard to remember what all the commands are, though, and for some tasks it takes longer, which is why we need a GUI in the first place - but no one fails to understand within the first few seconds what the PURPOSE of a command line is.

Look at your desktop - your real, physical desktop. Are there any icons on it? Is there a menu at the top of that fax you're holding? OK, there may well be a volume control knob on your desktop (presumably attached to some sort of sound system), but there sure as hell ain't no button to make that pile of papers disappear.

KDE and Gnome are doing a fine job of emulating, and in many ways improving on the current fundamentally-flawed GUI framework as manifested by Micros~1 and Apple (the most grievous exception being performance). They can still make more of a dent in the Linux learning curve, though, if they do two things:

- Enforce strong, clear standards for visual display of information. The method of enforcement and the particulars of the standard are obviously the sticking points; what open source developer wants to have someone else's way of doing things forced on him? But it can be done in such a way that it makes the developer's job easier, through improvements in the library.

- Provide more feedback. Feedback to the user is crucial for his understanding of the metaphor, whether that metaphor is clear or confused. That means instant response when you double-click on an icon - no waiting for the window to pop up, not knowing if your action worked or not. That means auditory responses that are built in by default. (If you don't like them, turn them off. That had better be easy to do, too.) That means - and this relates to the first change - enforce changes to the cursor's appearance when it moves over certain important areas.* That means that when you click and drag something around you either immediately get a message that you can't do it (in the form of a cursor change, let's not annoy the user with popups), or you immediately get the response that the user was EXPECTING. Who's to say you can't drag icons around the menus like you do on the desktop?

* BTW, this does not contradict what I was saying earlier about the problems with the metaphor of the mouse. If the mouse has to become a million different things - and, unless we scrap the whole GUI framework and start from scratch, it does - then it had damn well better let the user know what the metaphor is at this moment.

GNUstep

IMHO, the best bet Linux has for the desktop in the long term is GNUstep [gnustep.org], which is working towards providing a free(speech) implementation of the OPENSTEP API. I'm sure anyone who has ever played with a NeXT machine will tell you how great the OPENSTEP API + Interface Builder are/were for application development. It is the same API behind MacOS X's "Cocoa" API, so with all the forthcoming commercial support for MacOS X, these same apps could be available on Linux with as little effort as "make".

The OPENSTEP API also addresses problems still facing GNOME and KDE, such as cut & paste, and the hideous font support inherited from X, among other things. While X's selection-based clipboard system is great for text, it makes no attempt to handle any other content, like images, sound, or formatted text. I don't know how GNOME/KDE apps handle this, but many aps have a private 'clipboard' which only functions within that program, preventing, say, cutting an image from the GIMP and pasting it into a WordPerfect document. The pasteboard in OPENSTEP, IIRC, provides a MIME-type-based board which all other OPENSTEP apps can access.

Also, the Display PostScript (DPS) system which OPENSTEP is built on makes sophisticated graphics output simple to implement, and also provides consistency between what comes out of the printer and what shows up on the screen (WYSIWYG ;-). On most Unices, a completely different font set is available for X programs than there is for the PostScript engine (be it GhostScript, DPS, or a printer), so it is difficult to get real WYSIWYG apps under normal X. Also, antialiasing could concievably be built into the DPS engine, and with all the OPENSTEP apps using DPS, you could get a very nice display.

When GNUstep shapes up to be a full OPENSTEP implementation, it will provide an elegant basis for both application users and developers. With Linux being the buzzword it is now, developers will probably move to the OPENSTEP API when they find they can produce the same program for both MacOS X, which is shaping up to be a big consumer OS, and Linux, which wants to be the same.

Re:Standard?

Neither are standard, meaning that applications written for one are not written for the other.

Interoperability is not the point here but uniformity is. It is easy to copy the good looks of apple but it is very hard to copy its uniformity. Uniformity makes a system easy to use since you can reuse knowledge about using one program in another one.

Having and enforcing user interface guidelines is the only alternative to having a single source for all your programs if you want to achieve uniformity. Both KDE and Gnome are trying to do the latter by providing all sort of apps (I think there's even a browser component in KDE, where have we seen this before?).

The lack of standards in this area will maintain the status quo of non uniformity of applications for linux. Some will work better on Gnome, others will integrate better with KDE, some will ignore either and some will try to be compatible with both (thus limiting themselves to what is implemented in both).

So the user inteface for linux is doomed to be a wannabee MAC/Windows rip off. It may aquire the looks but is currently far from adopting the same uniformity.

Re:Standard?

Here's what *you* want (AFAICT):

A standard widget set used across applications on Linux.

You're never going to get that.

But it is a problem, so here's the way I've always said it could be implemented:

Make a library that has the requisite function calls (call them widget bindings or whatever) for each widget set. It then uses them to draw the application. It draws the application by checking to see what "theme" you have set up, and using those widgets to draw it. Common themes would of course be "Motif", "Xaw3d", "GTK", "Qt", "W1ND0WZ", or whatever your little heart desires.

Think of this as a widget API thunking layer. This project would of course be a lot of work, but the payoffs would be just as great.

And you'd want to build everything on your system with this, and have it dynamically linked. (You'd want them static either until it becomes standard, or if you're building a stand-alone app and don't trust library versions)

Then everything would look the same, (or different, if you gave different apps different themes--use X resources or something for that) there would still be many APIs to call, (pick your favorite, the user won't know the difference... mixing and matching might be bad, though) but you wouldn't have to include the whole GUI in your app (unless you're paranoid). Oh, and I'm talking about visual appearances. Standardizing layouts for users would be nice too, but let's ponder that some other time, ok? :)

Questions? Comments? Please?
---
pb Reply or e-mail; don't vaguely moderate [152.7.41.11].

Re:No, X does blow (Was:I saw a ploughman)

Who cares about colorspace management?

Wierd people like graphics professionals

I dont. Color displays correctly, thats all I care about.

Just because you don't use it or don't understand why it's useful does not mean that others can't/wont benefit from it.

Thats something else X does better then Windows or MacOS, it allows you to use different font engine servers, over a network even.

Oh my oh my, this is a stunningly ignorant comment. The advantage of font management under MacOS and Windows, that you seem unable to grok, is the fact that WYSIWYG printing is actually possible. That is, if you can display a font, you can also print it with the same methods. This is simply not true on Linux. ( QT tries to do this, but unsuccesfully. It tries to guess printer font names from the screen font names, but this is very hit and miss, and requires the user to configure their ghostscript fontmap or their printer / PPD ) In general, you cannot print fonts for a few reasons -- some of the fonts are bitmaps, which cannot be scaled to printer resolutions. There are more problems though -- X does not give the developer a way of retrieving outline files and metrics for the fonts. By the time font data reaches the X client, its already simply bitmap data. This means that there is no way for a developer to print screen fonts reliably. In case it's not blatantly obvious at this stage, let me inform you that this means that in general, WYSIWYG printing is nontrivial and involves a lot of ugly kludges. Applixware, Star Office, and Abiword all resort to ugly kludges to handle fonts. It appears that Koffice, which seems to use QTs font handling functions simply cannot print reliably.

Where's the unified print/display model? Its called PostScript....

No, this is not built into the display, at least not in older XFree86 releases. Consequently, it's not built into the toolkits. You've obviously never tried to program any WYSIWYG apps, because if you did, you'd understand how trivial WYSIWYG printing isn't. Tell me this -- how do you write a program, using just the X APIs, that displays a font, and prints that same font ? Once you can do this, you can come back and talk to us about "unified print/display models".

They seem quite usable to me and many other people, who are you to say they aren't?

Again this stupid and childish attitude. You're telling users that if they want these features, they're idiots and they should go away. But who are you to say this ? Are you a developer ? ( it doesn't sound like you understand any of the issues regarding developing GUI apps on UNIX, quite frankly )

Re:The Font HOWTO

The FDU HOWTO is cool, but too narrow in scope -- it is too TrueType centered , and acts as though TrueType is the only game in town. Check out the Font HOWTO [rutgers.edu] for something more comprehensive.