KDE enjoys immense market share compared to the marginal, poorly-maintained GNOME. And KDE provides what users want: 47 separate setup wizards which each must be completed before the desktop will be displayed, and applications which have menus for each of "Settings", "Preferences", "Configuration" and "Options"!

If you like the gargantuan KDE configuration system, then by all means use KDE. But in terms of usability by the general population, fewer preferences tends to be better. Joe Sixpack doesn't know or care about most of those things; he wants to change the theme, change the wallpaper and get busy doing the things he does with his computer.

Which, really, is the difference in focus between GNOME and KDE: GNOME aims for the general population, KDE aims for geeks who want to spend months tricking out their desktops.

If the sane defaults provided by your distro are good enough that you don't have to change them, then why aren't you whining that your distro's KDE is broken and dumbed-down?

Also, all the configuration options are still right there in GConf, and accessible via a GUI application that ships with GNOME and is in the menu system by default. And several of the options people use the gtweakui app for have become more configurable in subsequent GNOME releases.

What this is about is not dictating "this is the one true way, and don't you dare change it". What this is about is offering a small, sane set of the preferences most likely to be changed by most users. "Power users" who want to trick out everything still have access to all the available prefs, but bombarding Joe Sixpack with hundreds of settings he's likely never to use is just a bad idea; it wastes his time, it's confusing, and it makes him more likely to break things.

And isn't as bloated and slow as Gnome (well, really, Nautilus).

Konq's the one that wants to be Windows Explorer, not Nautilus.

And doesn't clutter my desktop with dozens of Windows whenever I want to drill down to something.

So turn off spatial mode, nobody's making you use it.

And can actually access files on the network.

I've never had a problem with this in GNOME. FTP, SFTP and other protocols work in Nautilus, and apps that use gnome-vfs can open files using those protocols as well. Maybe you should, I don't know, actually know what you're talking about before you spout off like an idiot?

Also: "imric doesn't like it" != "is completely broken is hated by all users, and will never be usable by anyone, anywhere". That's the problem so many geeks have when it comes to understanding usability; they can't get over themselves long enough to realize that other people might have different needs and expectations of a system.

Your entire argument, condensed from your posts, seems to be as follows:

\r\n
1. GNOME developers are hateful, spiteful people who would like nothing better than to stand over users' shoulders, waiting to slap their wrists with a yardstick if they even think of deviating from GNOME's default settings, which are the One True Holy Way.
\r\n
2. Nautilus sucks.
\r\n

Now, as I've attempted to point out, the first part of that is so hilariously out of line with reality that the only possible conclusion I can draw is that you're using your head to self-administer a proctology exam. I've pointed out that every single configuration option is still right there in GNOME, right there where all the geeks can get at them. I've pointed out that configuration options for which people devised popular third-party tools have been made easier to change in succeeding stock versions of GNOME. I've pointed out that the primary reason for GNOME's simplified preferences menus is that most people have no need of all of those things; they mostly only want to change a few things about how their desktop looks, and then get on with life. Which is exactly what GNOME does well.

And with regard to the second point, you don't seem to have any argument apart from repeating your opinion that Nautilus is "slow and unreliable". In my experience it's quite snappy, and I've not seen it crash since my last foray into GNOME (around eight months ago, so on GNOME 2.10; my previous foray into GNOME was some years previously, before Nautilus was integrated into the desktop). So you have your opinion, and you're entitled to it, but if all you're going to do is repeat "Nautilus is slow and unreliable", then I'll just stoop to your level and reply endlessly with "your momma's fat".

As for how you apparently hate spatial Nautilus: spatial Nautilus is the default, and, believe it or not, works well for many people. Anecdotally, I can point out that I like it quite a bit. If research found that the majority of users (note here that "majority of users" != "imric") had a better experience with the "browser" mode, then I'd be all in favor of making that the default; it's still easy for users who prefer something else to change this behavior to suit.

And again, a large part of usability, a part which doesn't seem to be understood by many geeks who spit ill-informed bile at projects like GNOME, is the ability to look outside your own particular use case and accept that other people may be different. But the pathological blinkeredness of most geeks is a tough phenomenon to overcome, so I won't hold my breath waiting for it to happen.

Configurability is not, by definition, bad. I'm sorry if your strange little worldview involves my proclaiming death to all configuration options, but I think even a cursory glance at my posts in this thread would reveal that such an assumption is delusional at best -- you've made up in your own mind a series of strawman "usability arguments" and knocked them down, and you don't seem to give a damn that they don't match up with reality at all. But I'll do my best here anyway, in the hope that perhaps your head can yet be extracted from your ass.

So, first principle: configuration is not, by definition, bad. But there is such a thing as too many options for a given group of users. In general, a particular group of users is not well served by configuration options if those options include a large number of items which that group of users is unlikely to know or understand, or if the number of items presented is so large or poorly-organized as to make it difficult for users to understand the structure of the application's configuration.

It is precisely in these ways that KDE fails. For example, it is not unheard-of for a KDE application to offer separate menu items titled "Configuration", "Options", "Settings" and "Preferences". These words are basically synonyms of each other, and to a user who is unfamiliar with the application, it is extremely difficult to intuit the actual differences between these items, or to accurately guess where a particular setting will be found. A better solution would be to eliminate the apparent redundancy of this system and collapse all configuration into a single menu item which encompasses all relevant configuration for the application.

Once a configuration dialog is presented, the problem of "too many options" often arises; KDE is designed by geeks, for geeks, and often bombards novice users with many more options than they need or want, and describes and labels them in ways which are not clear to users who are not "geeks". Most users would be better served by determining which options are most often used by the largest number of users, and reducing the default set of preferences to either that set or a slightly-larger superset of preferences.

As an example let's consider Quanta, which is often heralded as one of the great KDE apps. I've just popped it open, and I go to change some settings. There's a "Settings" menu, which is encouraging, but when I click it I find I have to choose between seven different configuration dialogs which Quanta offers, some of which are redundant and/or badly named. For example, there's a "Toolbars" sub-menu under "Settings" and a "Configure Toolbars" menu item. There's a "Configure Editor" item and a "Configure Quanta" item. But Quanta describes itself as an editor, so why do those need to be different?

Clicking on "Configure Editor", I'm presented with a configuration dialog. It has ten sub-panes, and the one that's immediately visible has four tabs, each of which is apparently applicable to one or more "schemas". Already, a simple HTML editor is beginning to rival Sendmail for complexity and ugliness of configuration, but I'm not done exploring yet.

Clicking around in the editor configuration, I find a number of options which even I don't see the need for, and I work in editors like this one for a living. For example, under the "Open/Save" sub-pane is a "Memory Usage" option where I can specify "maximum blocks open per file". This says to me Quanta's developer is too lazy to learn how to do efficient memory allocation, and would rather force me to figure it out. But what if I don't know? How am I supposed to figure this out if the guy who wrote the program can't? The answer, of course, is that this is a preference that has no business existing within the application.

On the same sub-pane is another option for "Folder Config File", where I can set "Search depth for config file". What the hell is a folder config file? How do I write one? I'm a "power user" and I've got no clue what this is about, so what use will this be to a novice? Answer: none whatsoever. It can be axed without any significant loss.

And that's just my immediate observations from looking at just one of ten sub-panes in just one of seven configuration menus in just one of the dozens of applications which make up KDE. The clear conclusion is that KDE's configuration system is horrifically organized and often runs to excess; options are presented in confusing ways which are difficult to reason out, and options are presented which have little to no relevance to the majority of the application's users and even at times have no relevance to "power users". Some options even betray poor coding practice on the part of the developers.

This does not mean that no configuration of any sort should be offered, or that configuration is somehow bad in and of itself. It means that configuration, as implemented in KDE, is utter shit. Now, get your head out of our ass and go to town.

Edit -> Preferences, "Behavior" tab, click "always open in browser windows".

Voila, old-school file browser with tree in the sidebar.

Now, you were saying? ;)

A project which never makes any mistakes (no such thing), or a project which makes mistakes and then doggedly refuses to fix them? Me, I'd rather have a project that fixes stuff when they realize it's not right.

You are cherry picking his points and ignoring the hard ones.

No, I'm first and foremost trying to avoid the inevitable conclusion some people seem to have entrenched in their heads, which leads to replies of the form "Oh, so you just think all configuration should be abolished in all applications, then."

The point is though, there are many trying to tell the GNOME devel team about those mistakes before hand, but are apparently talking to the hand that programs.

Perhaps it feels that way, but I don't think that's reality. Most usability professionals will tell you that, 99% of the time, ignoring feature requests and requests for changes is the only way to keep things working -- people either don't know what they want (most average users), or tend to request things which would only be useful to a tiny percentage of users ("geeks"). The only methodology which works is to start with a trained usability expert, design according to best practices, do user testing, and go back to step 1.

And in many ways, usability suggestions are subject to the same criteria for \r\n[link|http://www.catb.org/~esr/faqs/smart-questions.html|smart posts and questions] as any other aspect of development. Writing to a development list with "well, I think application foo ought to allow configuration of bar" is next to useless; what's needed is more like "in a significant percentage of use cases, we found that our users had trouble with application foo due to the unavailability of configuration for bar; here are some suggestions from our daily use for how that configuration could be accomplished in a simple and unobtrusive way."

And even then care needs to be taken by developers in what they actually decide to include; I don't see many people complaining that Firefox is "crippled" or "dumbed down", but they went through the same process GNOME is going through right now -- they had a huge, bloated, everything-and-the-kitchen-sink suite of applications, and pared it down to the feature set and options that most people want most of the time. Also, I find it amusing that people complain about things like gtweakui and say that third-party applications shouldn't be needed to perform whatever their pet configuration task is, but have no issue with using third-party extensions to add configurability (that is, configurability other than about:config) or features to Firefox. There's a double standard there, I'm thinking.

And for the record, I work in GNOME from time to time, but spend most of my time in Enlightenment, and have for years. I love GNOME now because I can just log in and get things done, but I also love the extreme configurability of E; I also recognize that there's absolutely no way to please all of the people with a single system, and accept that a system which aims for the general populace will occasionally conflict with the level of fine-grained control I'd like to have.

Before doing anything else, get over yourself. My reply was polite and in no way attempted to insinuate anything about you, personally; I was simply making a point about the types of "usability" requests mailing lists often get, versus the types of requests which would actually do some good. Meanwhile, you're the one who actually resorted to personal insults.

While you're working through those issues, I'll be waiting here patiently to find out exactly what the questions are that I'm supposedly avoiding, and what this "inevitable conclusion" is.

I raised the idea that some people thought there was an "inevitable conclusion", and then pointed out why I feel that it isn't. If someone has an actual argument for why it is, I sure haven't seen it yet, nor have I seen any other "inevitable conclusions" mentioned, so I'm really curious.

And again: what are all these points people are making that I'm apparently ignoring? Someone point me to one of them, I'm really starting to wonder...

I think that if KDE and its advocates really, truly want to aim for a broad userbase of varying technical abilities, then they've got a heck of a lot of work to do on usability, and their attitudes and implementations of configuration are just one aspect of that. This is, however, a noticeably different position from wanting them to "crumble into dust". And if some people like KDE as it is right now, I'm not doing anything to try to convert them; I'm just pointing out that the people who like KDE as it is right now tend to be almost exclusively of the "power user" class and so are unfazed by what would be, to less-savvy users, a mess of usability problems.

And as for the rest, I'm begging. I'm pleading. I'm on me knees grovelling for someone to show me even one of the points I'm allegedly ignoring or glossing over here. Everything I've posted in this thread has been met with one of two responses:

\r\n
1. "Oh, so you don't think anything should be configurable at all, then."
\r\n
2. "Yeah, well, somebody else made a whole bunch of good points which I won't mention or direct you to, but you ignored them all!"
\r\n

This does not a discussion make, I'm afraid, so it looks like all we've got is a good old-fashioned flame war. I don't generally like those, but I could probably get into the swing of it if that's what people really want.

Unless otherwise noted, the pronoun "you" is used below to refer to the original author of the snippet to which I am responding.

the red herring of characterizing having the ability to configure as 'configuring endlessly...' while characterising the DELIBERATE INABILITY to do fine configuration as a '...desktop with proper usability' is rationalisation, pure and simple. It's a pretty irrational rationalisation, too.

Having lots of configuration does not necessarily mean that users will endlessly configure the application. However, when I installed KDE on this machine, I had to spend a good hour on configuration before it was in anything resembling a state I could use effectively. GNOME, on the other hand, rather than pop up wizards when I log in and leave things in a fairly useless state to start off with, has sane defaults which mean I can pretty much log in and get down to business.

Also, GNOME continues to offer access to what you call "fine configuration options". It's simply done through GConf. Please do not attempt to misrepresent this in the future.

What you (and the Gnomes, apparently) don't get is that having the option to configure something doesn't meant it will be used by the rank-and-file. The idea that nailing settings down is even desireable is blown out of the water by the swift appearance of configuration utilities like tweakui.

What you and many others in this thread don't get is that while a very small percentage of competent users do want to have absolutely everything configurabel to the nth degree, this is not a capability that is desired by the remaining large majority of computer users, and furthermore, the structure required to present this level of configuration has a negative impact on usability for that large majority, in the form of:

\r\n
1. Added confusion as they attempt to decipher configuration items with which they're not familiar; what average user ever needs to configure the focus model used by the mouse, for example?
\r\n
2. Added difficulty and time spent trying to find the options they do want to configure; when there are twenty configuration options, finding the one you want is easy. When there are 200, not so much.
\r\n
3. Increased likelihood of accidentally misconfiguring something, due to a combination of the effects of (1) and (2) above.
\r\n

As for tweakui, I would again point out that this is factually incorrect; configuration options in GNOME are not "nailed down". Additionally, the GNOME version of tweakui does not offer very many options, and at least a couple of those options have since been exposed more prominently in GNOME's own UI. And finally, the desire of some users to have more configuration options does not mean that those configuration options automatically should be exposed to the majority of users; given the points above about the negative impact too many configuration options can have for most users, I'm quite happy to have them in GConf, and have third-party applications which provide hooks to that for those who do not wish to use GConf.

If your point had any muscle behind it, there WOULD BE NO TWEAKUI FOR GNOME.

I'm not really certain what the logic is behind your saying this; my point all along has involved the fact that different groups of users have different needs and desires. Arguing that no user should ever have a motivation to obtain a utility like tweakui is, essentially, arguing that the desktop should please 100% of users 100% of the time, a patent impossibility. When you have an actual realistic goal to aim for (like, say, providing a usable desktop with a minimum of fuss to a majority of users, something which GNOME currently aims to do and, to my mind, mostly accomplishes), I'll be listening.

I can go into the Registry on Windows too, to make changes! And Look! It Has A Gui App, Too - regedit.

I have never once claimed that GConf is the ideal solution to this problem, merely that it is a better solution than exposing a bewildering array of configuation to all end users. But then, nobody else in this thread has had any problem putting words in my mouth, so why should I expect you to act differently?

How do you then move from that position to "removing most configuration as done in GNOME" is good?

Now who's ignoring the other side's points? I've been pressured about this from almost the first post in the thread, and I've now made the case for reduced configuration several times. You've got some reading to do.

I would prefer a GNOME with all of the configuration options available in a non-confusing way. Or are you also postulating that this is impossible?

Pretty much, yes.

I don't think removing options make programs simpler to use... I think it makes them simpler to configure, which is different. And since configuration is something typically done once, that's a bad optimization.

Simpler configuration does mean simpler use; the fewer the number of items which must be confifgured before use, the easier it is to begin getting Real Work done. Also, plenty of configuration tasks are performed more than once; look at wallpaper and desktop theme choices for a couple good examples.

I think trying to focus someone's use of an application is a bad thing to do. Make common things simple and easy, but don't make hard things impossible.

So... the most commonly used options should presented in a prominent "Preferences" system, while less-used or "more complicated" options should be presented via a separate mechanism which is not as prominent? How is that different from what GNOME does right now?

The fact that they fixed it later on rather points to the silliness of the choice in the first place, doesn't it? And by extension, perhaps, they might have made other silly choices?

This is, or quickly becomes, a red herring. The subject of debate is the reduced number of configuration options GNOME exposes in a prominent fashion, not the default options set for a particular application or applications, nor whether a particular option should be exposed in the "main" preferences system or only in GConf. I would, however, argue that the particular choice to make it "difficult" to switch Nautilus to browser mode was incorrect, though since I don't know what information the developers based that choice on, I can't say too much more about it. The fact that a particular choice was made incorrectly does not, however, provide any support for the KDE style of configuration; it merely provides support for an argument that the methodology used to make this choice should be examined, and corrections should be made to the development process to avoid a repetition.

Right now, I have to setup launchers for these types of things. WTF, now I have drawers full of lauchers for my specific tasks... yeah it is manageable, but a PITA. I want the window manager/etc to see the stuff I do and if it matches a rule... of it does things.

Again, "I want application foo to provide bar" does not automatically translate to "application foo should provide bar"; the choice of whether to add a feature or configuration option must be based on careful analysis of the costs and benefits of doing so. Also, there are plenty of window managers available which will do what you want, so complaining that one in particular does not seems, to me, a bit odd. Again, no desktop or WM can or should attempt to please 100% of users 100% of the time, as that is an impossibility.

being able to enable snap-to in metacity (or magnatism/edge resistance/etc) so I can easily maximize my use of my workspace real-estate without overlap. [...]These are the kinds of things I want back. For now, I have a launcher farm. I can only hope.

Same reply as above.

So what's the usability model again?

This was part of a point about how Nautilus made it "too difficult" to switch to browser mode. That point turned out to no longer be accurate, and in any case I've already responded to it above.

The remaining items quoted were people accusing me of "cherry-picking", or ignoring points made by others and of setting up straw men. I thank Greg for actually taking the time to tell me what it is I've supposedly ignored, but beyond this I'm afraid I'm still somewhat in the dark. As I saw it, it was the people who jumped to "Oh, you hate all configuration and think it should all be abolished" who were attempting to knock down arguments which were not actually being made. But hey, that's just me.

GConf is a 'registry editor'. It is the same thing as regedit. It sucks for configuration. Your elitist attitude is showing.

And I have said, at least twice now that I can recall in this thread, that I don't feel it's an ideal solution. I do feel it's better than "expose all options to all users all of the time", however.

This makes no sense. If you are relying on so-called sensible defaults, the that structure has no impact on your "usability for that large majority". Making things a PITA for any of the remaining minority that may want to customise or support Gnome is not a helpful.. In short, it is a net minus for the usability of the desktop.

OK, Interface Design 101 time. Each and every time you add some new element to an application's interface, even if it's "just a checkbox" or "just a toggle", it will have a certain impact on a certain number of users. A sort of usability calculus must then ensue, often involving one or more rounds of testing with actual users, to determine whether the number of users negatively impacted, and the amount of the negative impact on those users, is smaller than the number of users positively impacted and the amount of the positive impact on those users. If this turns out to be "no", then the new interface element generally should not be included.

And as I have repeatedly provided reasons why exposing large numbers of often-esoteric configuration options to all end users often results in a significant negative impact for the majority of users, I find myself unabel to make sense of the rest of your argument here.

Nobody has to do that, you do it yourself quite eloquently. I was responding to what you think are 'points', nobody made up anything about wwhat you were trying to say.

I began this thread with some semi-joking comments about KDE. It quickly devolved into other people's assumptions that I am against configuration of anything, for any reason, at any time, and their attempts to belittle such a position. My own efforts to point out that this was not, in fact, my position and that I did not, in fact, advocate it seem to have gotten lost in people's knee-jerk reactions and exclamations that I was not responding to their points. I found this rather odd, as I tend not to respond to a strawman as if I held the position its creator claimed I did, because this tends to cement in the mind of its creator that his strawman is in fact a correct representation of my position. However, I chose to indulge and attempted to steer the conversation away from such hand-waving and fallacy, and in return was met with rudeness and flames.

So. Nice bunch of folks you've got here, huh?

You simply have no idea what HCI models entail and why it is so important to get it right...which Gnome (as many have pointed out) was close to and then moved away from as a paradigm.

Then pray enlighten me, because UI design is a fair chunk of what I do for a living, and while I don't consider myself an expert by any stretch of the imagination, most everything I've ever read on good UI design indicates that, the simpler you can keep an application in terms of providing sane defaults and avoiding the need for much end-user configuration, the better, and that "everything should be configurable in every imaginable way" as some models, KDE's in particular, advocate, is the sort of thing which leads to overly-complicated and unusable systems.

So please, do let me know how I can correct my woeful ignorance.

GConf is not a settings manager that "joe user" will use.

I never said it was. And yet I'm the one being flamed in another thread right now for an apparent lack of English comprehension.

And contrary to what you believe..."joe user" wants to be able to personalize the experience.

I have never stated otherwise. I have merely stated that there are a huge number of options which "geeks" and "power users" tend to want and use, which "Joe User" neither knows about, cares nor understands. "Joe User" doesn't know what a mouse focus model is, for example, so why should he have to sift through that in his desktop environment's configuration? For all but an extremely tiny minority of users, "focus follows click" is both expected and desired, so why does this need to be exposed in an environment's primary configuration system?

You are also completely off the mark about KDE...and before you confuse me with someone who has a stake in this dispute...I use xfce4 as my default window manager. It does what I need and doesn't do much else...which makes it less of a pig than Gnome and KDE.

And for the most part, I use Enlightenment. My comments about experience with KDE are based on distributions which do not provide it by default and thus with getting a "sotck" install with little to nothing pre-configured (as opposed to KDE-centric distributions which would obviously take some steps to alleviate this). However, the required set of "sensible defaults" seems to me to be a bit large to have to configure if one is using just a "stock" KDE install. The initial setup wizard is some help, but I tend ot feel that if a desktop environment requires a setup wizard at all, something is likely wrong with it from a usability standpoint.

Simply put, to me if a TeakUI tool is necessary, the application is broken. If it was correct, there would be no need to develop >work-around< tools that (alas) turn into "killer apps".

I've said now, several times, that the GNOME tweakui has, as far as I can tell, led to changes in which options are exposed by default in GNOME's primary preferences system. This is an example of developers learning to improve through user feedback. However, the existence of a tweakui for GNOME has been used in this thread as an argument for the position all every conceivable configuration option should be exposed to all types of users at all times, a conclusion which does not follow logically from its stated premise and which, as I have been trying to say since the beginning, almost inevitably creates a worse user experience than GNOME currently offers.

You dismiss this by accusing the group of being techno-geeks and not understanding that to make it usable you must oversimplify...and us geeks will just have to use our geek-fu to (now understand this statement carefully) reclaim lost functionality that has been removed to (in your words) improve the experience.

Not quite. I've been pointing out that the functionality is still accessible, yes, but one of my earliest posts in this thread was "nobody's forcing you to use GNOME". If you want something which exposes tons of configuration all the time, then use something which does. But don't go to a project whose aim is a simple, usable desktop for a broad userbase and ask for that, because it's not really compatible with that aim.

and adding a click behavior option is a line and a couple of radio buttons in a screen that is already there

Every UI change, even when it's "just a couple radio buttons", has an impact on usability. Every such change must, therefore, be rigorously determined to be necessary. As I've said before, many people here seem to have the attitude "how can you justify removing this?" while my attitude is "can you justify keeping it?" Applications should have as many configuration options as are absolutely necessary, and no more, because when you start throwing in edge cases like "well, maybe one or two people would want this to be configurable like that, so let's just go ahead and add it", you start down the road to feature creep and poor usability.

Dumbing down the interface for everyday users isn't going to win many friends and influence many people that use Linux.

I really, really, really hate the phrase "dumbing down". That's how most "geeks" perceive it, unfortunately, but it's very sad that they do because it implies "those average people are too stupid to use our l33t system!" In reality, usability has nothing to do with "smart" and "stupid" people, and the best interfaces are those which are intelligent, not those which are "dumbed down". But any attempt to simplify the use of Linux applications, or remove unnecessary hurdles to learning to used Linux, is immediately set upon as "dumbing down", even if it results in interfaces which are more intelligently designed.

Case in point: I'm helping to develop an application which will profile certain types of information about web sites. As part of its configuration, it needs to know things like whether a site has a search function and what URL and URL parameters it uses in searches. Comparable applications tend to have an input box for "search URL" and another for "search query delimiter", along with explanations (sometimes lengthy explanations) of what these mean. My proposal for interface to implement this is somewhat different: since a large part of our target audience is bloggers, I proposed that it instead ask "what blogging software do you use", and offer a list of Movable Type, Wordpress, etc. Since those packages each have their own standard search URLs and parameters, that information can be inferred from the blogging software without needing to mess about with asking a user for it. Of course, an additional "I don't see my software listed here" option would pop up a traditional search URL/query parameters input, but for the vast majority of cases we've just avoided the need for that.

This would be called "dumbing down" by many, but what it really involves is making the interface more intelligent; it doesn't need to ask potentially-confusing technical questions because it can usually infer the answers it needs from a simpler question.

(It also, you'll note, removes at least one item of configuration. I'm a fan of that.)

I'd never recommend anything try to be "more like Windows", for any reason, least of all a perceived winning of new users. I think Linux usability can learn a lot from the Mac, though -- there are many things that Apple has learned to get just right over the years (and, on the flipside, many things they've horribly fucked up; nobody's perfect), and the perception that "we don't do things that way" would hopefully be somewhat lessened by the fact that the Mac OS, too, is built on top of UNIX.

And that essay seems, to me, to be verging on NIH syndrome. Not a good thing.

I prefer to use a distro that's *not* oriented toward a particular desktop when trying it out; I'm sure Kubuntu and others provide nicely preconfigured stuff, but from a usability perspective I'm interested in the stock shipping version of the environment, not the version you get after a third party intercepts it and fixes things to make it easier ;)

And just to drive home the point that I've been making about different types of users wanting different things: I use Enlightenment by choice, and it requires more configuration than any three other desktops/WMs put together. I like that sort of fine-grained control. But I know that not everyone does, and in fact the vast majority of people wouldn't have a clue what to do with such a thing. Which is why I think anything which aims at a broad userbase (as GNOME does) should work on simplicity and sane defaults rather than "everything is configurable in eighty zillion ways", because the former approach nets many more users than the latter.

There are times, many times, and not all times, when I personally like the level of configurability Enlightenment provides. However, I understand that not everyone prefers this level of configurability, and that many people would find the idea of having to spend as much time tweaking a desktop as I've spent on E to be absurd. For these people, GNOME exists; it's very easy to change the things people most commonly want to change, and it just gets out of the way and lets you Do Stuff. There are times when I'm a fan of that myself; the only things I've ever changed in my GNOME setup are:

\r\n
• The theme.
\r\n
• The wallpaper.
\r\n
• The fonts.
\r\n
• The set of application launchers and applets in my panel.
\r\n

These things are all extremely easy to do in GNOME, and don't by any means represent the extent to which GNOME is configurable. But this leads to a useful observation: not everyone wants or needs a high level of fine-grained configurability, and often a desktop which does not offer that is just as usable, possibly more so, than a desktop which does.

I don't think a file manager is a great example here, because it's expected to be something of a jack of all trades. But right now one of my favorite applications is [link|http://basecamphq.com/|Basecamp], a web-based project-management system with an almost absurdly small number of configuration options. The developers started with a single thesis: project management is not about graphs or charts or four levels of committees signing off before something can go forward. It's about communication. So they built an application which made communication simple and easy, and kept things tightly focused. At first they only used it in-house for their web-design clients, but then they began marketing it as a service and now are making boatloads of money from it. And its focus hasn't hindered use one bit -- plenty of web designers use it for project managemnet, yes, but it turned out that "project management" is a pretty broad thing, and now they've got customers using it for everything from major corporate processes to wedding planning.

So "focus" doesn't mean "define exactly the set of things users will be allowed to do, and forbid all other uses".

Configurability, in and of itself, is not an evil, should not be unilaterally abolished, does not cause applications to lose their focus, does not cause global warming, and did not shoot JFK. And I don't care who or what they are, I will flame the everliving hell out of the next person who claims I said it does.

TOO MUCH configurability, and POORLY-ORGANIZED and POORLY-THOUGHT-OUT configurability, on the other hand, have all sorts of negative impacts on an application.

With regards to focus, over-configurability has a nasty way of causing feature creep; "well, we want people to configure this, so we should offer a way to configure that and those as well, and that means we need to support option X so we'll have to refactor foo..."

And in terms of user focus, it also has a nasty habit of distracting attention from what the user is actually trying to do; there are so many applications out there which immediately pop up configuration dialogs first thing after launching, when so many of them don't actually need to do that, and would run just fine for most users with a sensibly-chosen set of defaults.