Slashdot Mirror
The Command Line - Best Newbie Interface?
An anonymous reader writes "This essay describes the surprising results of a brief trial with a group of new computer users about the relative ease of the command line interface versus the GUIs now omnipresent in computer interfaces. It comes from practical experience I have of teaching computing to complete beginners or newbies as computer power-users often term them."
I'd wager that computer literacy amongst people who've tried Linux would be twice what it is today if when you typed help foobar bash would perform a man foobar if 'foobar' wasn't a builtin command. And it'd probably be double that if you incorporated some kind of search facility too. Type in help disk space and get a hit on the df command, for instance.
And isn't it nice that Mac OS X now gives you the best of both worlds :)
I find it amazing how many computer "experts" are dead in the water when the mouse doesn't work or the GUI doesn't come up as expected.
Too bad only the "old-timers" seem to appreciate the power of the keyboard.
To Terminate, or not to Terminate, that's the question - SCSIROB
It seems to me that if you start someone on a command line versus starting them on a GUI they learn a little slower but acquire a deeper understanding of the computer.
point-and-click and drag-and-drop don't encourage any actual understanding of ways in which a computer interprets commands.
lysergically yours
Apprentice: "What is that, Master?"
Master: "It's a command line. The instrument of a Unix Programmer. Not as random or clumsy as a GUI. An elegant interface for a more civilized age. Before the dark times. Before...Microsoft!"
"Backups are for wimps. Real men upload their data to an FTP site and have everyone else mirror it." -- Linus Torvalds
I believe it is because the command line is very simple and compact, single-threaded and simple to understand.
Modern GUIs have many things going on at once, which can confuse people who have no idea what is going on. Windows pop up, there are icons to deal with, they have to search through endless menus to find what they want.
The command line however has simple command to remember instead of complicated graphical procedures, and the status of the command line really never changes. If you mess something up, you are still back where you started, but in the GUI, a user could close a window to open another one which obviously confused people who don't understand what they mean.
If only there was a -1: Author's Girlfriend moderation option...
Rich
But you think a command line would make using a digicam easier? A microwave? A thermostat?
The computer as a non-specific device is a fundamentally flawed (though useful) contraption. The command line, GUI, and other UI creations are all hacks to help users get around the problem of genericity of the machine.
As computers get more powerful and more 'intelligent', computer user interfaces like these will wither away and something more straightforward like controls for a camera, microwave, or thermostat become the primary UI of the computer. This means that innovations in computer operating system design must be made so that the OS can guess what the user wants to do and present an appropriate, simple interface.
I really look forward to the day when the concept of the PC disappears.
I have been pwned because my
When she worked she worked with DOS based programs. I guess now these are so much easier to understand because you actually "talk" to the computer albeit through a keyboard and with a very limited command set. Maybe the mouse driven GUI is a bad inbetween step from the keyboard-only days to a time when computers understand conversation.
One of the things I really miss when I sit behind a windows computer is a bash shell, tab completion, gcc, vi... and you usually arent allowed to install cygwin on people's systems :)
Of those to whom much is given, much is required.
...as part of its Win95 hype machine. Microsoft likes to point out that pointy-clicky is sooo much easier than the "arcane" and "cryptic" command prompt, and tried as hard as possible to hide it. Microsoft certainly didn't try to improve its command prompt much, and even in modern version of Windows it still retains a lot of its retardedness inherited from the early days of DOS.
The question is, why? Sure, newbies hate it, but it's really useful to have a powerful command prompt, so it wouldn't hurt to include it. Even Macs have them now. Windows would be much more tolerable if it had a Unix-style command shell out of the box. Yet Microsoft feels the command prompt should die and it seems (at least from my point of view) that it's included only grudgingly in the OS.
"Backups are for wimps. Real men upload their data to an FTP site and have everyone else mirror it." -- Linus Torvalds
Remember??
I still today spend way more time in a 4DOS (or well, 4NT) prompt than I do in the windows GUI.
Virtually everything I do at work is done via 4NT BTM files (batchfiles that only work in 4DOS/4NT).
The CMD line interface is really good, and it makes me a little sad that new computerusers don't realize this.
Some things are just done so much quicker using the CMD interface...
For example, I use multiple computers at work, and I have to transfer files between them all the time.. I do this using 4NT BTM's, and it takes a few seconds to transfer a file... doing the same thing in a FTP app takes much longer...
Now that's newbie friendly.
READY.
#
Configure the DNS server address staticly:
/etc/resolv.conf
/etc/ directory. If any of these configs get corrupted you can edit them manually to fix them, or copy originals from packages, or copy them from another source and edi them to fit your needs.
In windows:
Goto start --> start --> control panel --> networking --> select the correct device --> options (or whatever I don't f-ing remember) --> select TCP/IP --> select options --> hit the DNS tab --> Select "I want to specify dns addresses bullet --> type in ip address in the boxes.
Networking specifications are spread throught 100's of dialogs and wizards. Some can set conflicting settings, some can override the settings of others. Some are grayed out to prevent these conflicts and you need to figure out why. All system wide configs are stored in the same place as program specific configs and user configs. If any program or installer corrupts this one binary file it will render the entire OS unusable.
In Linux (if the dns server is 192.168.1.1:
echo "nameserver 192.168.1.1" >>
Remember resolv.conf because DNS is to resolv the URL names into ip addresses. System wide configs are stored in plain text files in
I did some teaching of command line for non-English-speaking students and the biggest obstacle was the language barrier. They couldn't remember the command and they didn't understand the output of the command. Even if they knew some English, there were always some technical terms they couldn't understand and they felt intimidated. This way they were much more efficient in a localized GUI.
Save the bandwidth. Don't use sigs!
from my web log, I think it's appropriate and strangely enough, quasi-religious...
We all strive to be big monolithic programs, with fancy buttons, big memory footprints, environments where people, if they want to do anything, must go through us. We strive to be pre-eminent on the desktop, world stage. We crave fame. Look at me we say. Look how important I have become. I am an Office Suite, hear me roar. Look how much I can do. If you want to do any work, you must come through me.
[snip]
We must teach our brethren the ways of the Unix shell, for if we don't we will forever be trapped handcuffed in that big shiny plastic bubble of modern life, where we see but we can't interact. We must go back, back to the beginning and learn the first lessons.
Toddlers are the stormtroopers of the Lord of Entropy.
Before, I was concentrating on the syntax of control structures. Like having:Rather than: I could think about adding in a better help system as well. I've got a few months left of design work.
And I need to fix the lexer, too. In a recent presentation, I found a rather embarrassing bug. The concatenation operator in my shell is the same as perl's, the full stop, or period, ".". Cleverly, the shell can also treat numbers as strings and strings as numbers.
Unfortunately, it was all a bit too clever.
The expression 3.0 + 2.0 was parsed as (("3" . "0") + 2) . "0"). Giving 320. Oops!
But given a little more work, maybe I could get it to solve some of the problems mentioned in the article above. Could be an interesting thing to do.
The pity is that GUI usability peaked sometime in the late eighties and has declined since then, as the rise of "computer literacy" has created an expectation that users will master complex UI's, and the rise in computing power has removed any barriers to marketing-driven featuritis.
The most telling point was the discussion of "discoverability." "Discoverable: The interface must, from first switch on, provide a clear direction for a new user to go. At each stage it should encourage experimentation while providing adequate notice of important or key features."
In the 1980s, GUIs were intentionally designed to facilitate discoverability were far more "discoverable" than CLIs of the day. They were also intended to be forgiving. The user was supposed to feel empowered to try things, confident that there was always an "undo" to bring them back. On the Mac in the 1980s, "Undo" was far more prevalent and worked far more consistently than in today's software, in which many operations commit you to something whose effects you may not understand.
As for "dialog," UI designers understood that well. Why do you suppose that what are now called "screens" were once called "dialog boxes?"
What the article is really saying is not that CLIs are better than GUIs, but that a) modern UIs are not catering to the needs of the average user, and that b) modern UIs have gotten so badly designed, cluttered, and complex that they have become less usable to beginners than CLIsbecause GUIs have deteriorated, while CLIs have benefitted from benign neglect.
"How to Do Nothing," kids activities, back in print!
Build it, and they will come^Hplain.
Having done everything from assembly language programing for embedded systems to UNIX CLI (Command Line Interface) to Macintosh, I find that CLI is distinctly inferior to GUI (Graphical User Interface) for all but a few tasks. I challenge CLI users to do any form of word processing from a shell prompt. Even the most hard-core of them will resort to vi or emacs which use a primitive pseudo-GUI (and yes you can create and I have used a pure CLI text editor, but it is extremely painful). I don't want to even think about trying to replicate Photoshop with a CLI.
CLI = Dialog? The article mentions the notion of CLI as a dialog. But this is a misleading metaphor because so many CLI commands create invisible effects. You tell the computer to do something and all that returns (in most cases) is a command prompt. At best its like teaching someone to to do a job while speaking through a door. You give a command to do something (e.g., move a file from directory to another) and then you have to give a command to see the results (ls).
Discoverability: GUIs also provides visibility on to the set of available commands and functions. By browsing through the menus (which are usually nicely organized), you can learn the functions of an application. In contrast, a CLI-only machines provides no obvious way to learn about commands that you did not know existed -- at best you can access an alphabet soup of cryptic vowelless cmmnd names and then access the man page on each command. Therefore, GUI applications tend to be self-documenting, CLI commands require that you first know of the existence of the command and then you must read the man pages (grepping the man pages sometimes works if you know the jargon for what you are looking for).
Undo command: Most well-behaved GUI applications further support user learning via experimentation by having an undo command (and a revert command). CLIs tend to be irrevocable with no possibility for undoing inadvertent damage by a novice user (short of reloading the entire machine from a backup). Its no wonder *nix people get upset at the thought of novices on computers. But this lack of an "undo" is the fault of *nix CLI (it could easily be remedied with automatic file version tracking and journalling).
GUI is the superset of a CLI: Some people complain that GUIs take too long and I agree with them. CLI does offer a faster interface for experienced users. Yet a good GUI offers keyboard shortcuts that let experienced users invoke commands from the keyboard. While it is easy to have a keyboard shortcut available and shown in a mouse-oriented graphical GUI menu, it is hard to have a graphical menu shortcut in a keyboard-oriented CLI. With a GUI that shows the shortcuts in the menus, the user can learn shortcuts as they use the machine. Thus, I would argue, a GUI is the superset of a CLI.
Direct Manipulation: And CLI's inferiority is not just a matter of the learning curve (although that is a big disadvantage of CLI). For some tasks, a direct-manipulation WYSIWYG GUI is vastly superior. I challenge *nix people to build a CLI-only version of Photoshop. This is more that a matter of ease-of-use it is a matter of creating a coordinated interface between a person and a machine. While CLI forces the user to preform a completely defined action (e.g., type in a command that turns pixel 100,103 in file x to RGB value 128,128,200), a GUI lets the user try something (select a paint brush tool from the toolbox, test it somewhere, undo, use the tool somewhere else, etc.)
Control Panels vs. Config Files: The article claims that modern GUI-driven OS have too many control panels ("To master modern GUIs, one must recall the operation, layout and relation to each other of hundreds, if not thousands, of such panels." Yet how is this same functionality attained in a CLI machine? Config files are the absolute worst because they offer no form of input checking or potential for embedded help. Most config f
Two wrongs don't make a right, but three lefts do.
I understand your frustration, but also disagree with it. Allow me to explain.
The concept you're bringing up is black boxing: a simple (or standardized) user-interface that hides a complex system.
The pros of black-boxing are obvious: black boxing makes it possible for many people to operate complex equipment (think cars, phones, computers). There is also a psychological factor at play here: as many people seem to suffer from self-learned helplessness when dealing with technology, black-boxing (e.g. think nice-and shiny i-POD, or automatic-transmission lady-car) helps them to overcome their physological aversion and acutally use the device.
Now, as already stated by another user, the main con of black-boxing is what happens when something goes wrong: black-boxing provides a very poor way of dealing with failure-conditions!. There is also another con, which is public perception: black-boxing makes people believe that the equipment they are operating is simple, while in reality it's very complex. But because they reason from their internal "simple make-believe model', they can make mistakes that they wouldn't have made, had they known a bit more about the equipment.
To provide an example: Here in Europe, most cars are manual transmission (typically 5 forward and 1 reverse). When driving 80 kph in a flat country like holland, it really doesn't matter that much whether you drive in 3rd gear of 5th gear. However, when some Dutch people drive through Switzerland during the summer holidays (with a big trailer behind the car), they persist in driving in 5th gear up-hill, their argument being: "Hey, don't worry man, the car maintains speed uphill, so what's the problem". Of course, had they known a bit more about how an engine works, they would have switched back to 4th (or even 3rd) gears, to allow the engine to run at a higher specific torque, and to allow the cooling fluids to be pumped around more quickly to cool the engine better. Our black-boxing people usually end up overheating their engine (and blocking traffic and creating big jams in the Alps, grumble! ;) You see: it pays to know at least a bit about what's under the black cover of your black box!
In the same way, your shiny XP thingy makes your computer seem a simple device. But of course, it is still a really complex device. We've all seen people do stupid things because they persisted in acting according to how they believed the computer works (their simple model), rather than basing their actions on how the computer actually works. Knowing a bit about the latter, however small, enables you to do better in times of failure, but also in general.
On a personal account: I learned C after I learning assembly. I experienced this as a big bonus, because after using asm, you know what pointers actually are and how they work. I've seen people program Java and doing stupid memory allocation things, because they had no clue about what happens when you do "new bla" or "delete bla".
To conclude: black boxing is ok, but you need to keep in mind that you are still operating a complex device!
Support a Europe-related section on Slashdot!
Well, I'm an "old timer" who does _not_ appreciate the command line.
By "old timer" meaning, admittedly, not 60 years old and having started on punched cards and electronic tubes.
I did, however, learn programming in hex, not even assembly, on a ZX-80 with 1K RAM. In 1K you didn't even have space to run an assembler. I had a big old paper notebook with 1 page per command and a matrix of the registers involved. E.g., this is the page for "ADD", take the column for "A" and the row for "B", and there you have the hex code for "ADD A, B".
I continued through such mis-haps as editing source files on a CP/M machine with a hex-based disk editor, because the PHB forgot to also give us a text editor. Sad, but true.
You know what? I _don't_ miss those days.
The command line is powerful and all, as long as you already know _exactly_ what you're doing. It is a pain in the donkey when you don't.
The time and effort to get past that learning curve is not fun, and not what Joe Average wants. Heck, it's not what _I_ want.
I do _not_ find it fun to spend hours digging through obsolete, incomplete man pages just to find out that I need to type some obscure command like "obscureProgram.sh -XFGXRmnq -i filename1 -o filename2 -c OBSCURE_COMMAND_CODE -p some_obscure_regexp -f unix-style-font-name" just to get something done. Bonus points if it expects me to already be in some directory, and some obscure configs to already be set right. More bonus points if it doesn't even do the whole job, but expects me to pipe it through other obscure programs. Double bonus if it outputs some cryptic error messages like "1962 Short School Bus", that don't even tell me what the **** went wrong there.
Gimme a break. My time is too valuable to spend it on that kind of crap.
Give me a GUI which has input fields for the stuff I need to enter. If it's a file name, give me a good file selector dialog, don't expect me to manually list directores 20 times to find it. If I'm supposed to enter options, give me checkboxes, radio buttons, or drop down combos. And, ffs, give me an up to date help for it. And clear, humanly understandable error codes.
And you know what? I'm surprised how much energy goes into defending the sacred right to produce crap code and piss-poor interfaces.
Here's an idea: if half the time that went into whining about how the 60's command line interfaces are better for the user, went instead into throwing together a simple user-friendly TK or ncurses or whatever GUI, we'd all be far better off than we are today.
And let me get back to the part where I've said "The command line is powerful and all, as long as you already know _exactly_ what you're doing. It is a pain in the donkey when you don't."
The problem there is that to get to know exactly what options to type there, you have to invest ludicrious ammounts of time into that. Which for most people isn't justified. If you'll configure printing on your home network maybe 4 times in your whole life, consider the following two situations:
1. spending 4 hours to learn how to do that with CUPS and only with command line tools. But then you can do that in 30 seconds flat each time.
2. spending 5 minutes each time doing the same in Windows, through the GUI
Believe it or not, solution 1 is _not_ an improvement. On the whole, the l33t Unix command line way took 4 hours from your life, while the point-and-drool Windows way took a total of 20 minutes. The winner is... the GUI.
Yearly millions of hours go into just learning to use some crap software. It isn't learning some l33t skillz, it isn't "getting a clue", it's just _wasted_. It's time when you're _not_ doing what you needed to do in the first place. Time where, like in the example above, instead of already having your file printed on that networked printer, you're still searching through obsolete man pages and trying stuff that fails for no obvious reason.
A polar bear is a cartesian bear after a coordinate transform.
Look, if the car starts bouncing, I know I have a flat tire. If the engine is having problems, a red light on my dashboard tells me to take the car in. When the car struggled to shift gears, I knew it was a transmission problem. The CAR informs me of what is needed and I take it to the specialist. The basic maintenance schedule is given to you when you buy your car, there are no surprises.
Regarding a license... you need a license because with an automobile because the misuse of an automobile can endanger others.
Look, I'm a computer engineer by training, know my way around Unix, know my way around NT (MCSE, CCA from old career), and know my way around a CLI. Guess what, when I want to put together a business model, I want a graphical spreadsheet that is out of my way. Do I know ALL the Excel functions? No, but I know the ones I use regularly, and the help system gets me through the rest.
I have no desire to memorize arcane commands, I want to build the business model. Will I learn arcane commands to speed things up? Sure, but on MY schedule, NOT yours. I buy the machine to let me get my work done, and I pay the premium for a laptop to do so on the road, and I pay the (admitted small these days) Apple premium to have wireless support that is amazing and to be able to have my Unix applications on the road. These are all my choices.
Guess what, I don't pay Open Source programmers, so they don't have to listen to me. However, don't complain that I don't have an open source desktop when you put out bizarre GUIs that interfere with my workflow.
My Mac stays out of my way and lets me do work, MUCH faster than Windows or Linux would let me. In the end, that keeps Linux a step away from "global domination" or whatever the goal of the week is.
My computer should work for me, not the other way around. Sure, I had to demonstrate an understanding of traffic laws to get on the road. But guess what, EVERY car I drive is nearly identical, and the controls don't change. That consistent UI makes cars a success.
Windows has problems, no question. But it is "good enough" for most, and it is pretty cheap.
For a single-purpose machine, Linux is fine, I can train my people on 1-2 applications... If GNUStep was further along, or Qt didn't blow as a RAD tool, then our dedicated personnel would be on Linux machines and not eMacs. However, if you think that Linux is good enough for a business power user, you're sadly mistaken.
The integration of the Office Suite has been AMAZING for a few years, and the combination of Word, Excel, and Powerpoint is TERRIFIC for business users. Until GNU can mimick that power (NOT mimick the widgets, but the power and integration) then it is NOT a viable desktop system for the modern business power user.
Ever tried telling people on phone how to do a task in Windows?
"Click Edit, Options, and... err, wait, yeah, Page settings, Margins..." (5 minutes later, or 10 minutes if you don't have the same program in front of you) "Wha? I told you to click on OK, not Cancel!" (10 minutes fly by...) "So it doesn't work? What do you mean there's a setting like that? Why didn't you say so? I'm deaf and stupid. Sorry."
As opposed to:
"Just type '/sbin/ifconfig ppp0'. ...Wha? that's 'slash-s-b-i-n-slash...'" and a moment later "So it isn't even connected. Right. Try typing 'pon'."
The GUIs may be good to work with, but CLI is easier to explain. Especially if you're used to doing GUI things in one way and others do it the other way.
GUI is like a religion - you "learn" it, and only you can explain what it means to you. CLI is like science - you learn it, you can explain it, and it means the same thing to everyone - except people who argue about "useless uses of cat(1)" and proper order of parameters, and whether to use sed(1) or perl(1), of course.
And now I need the freaking coffee already.
RPM is supposed to be easy to use... but the dang man page is so long... Here's just the SYNOPSIS section at the beginning:
...
...
...
...
...
...
...
...
NAME
rpm - RPM Package Manager
SYNOPSIS
QUERYING AND VERIFYING PACKAGES:
rpm {-q|--query} [select-options] [query-options]
rpm {-V|--verify} [select-options] [verify-options]
rpm --import PUBKEY
rpm {-K|--checksig} [--nosignature] [--nodigest]
PACKAGE_FILE
INSTALLING, UPGRADING, AND REMOVING PACKAGES:
rpm {-i|--install} [install-options] PACKAGE_FILE
rpm {-U|--upgrade} [install-options] PACKAGE_FILE
rpm {-F|--freshen} [install-options] PACKAGE_FILE
rpm {-e|--erase} [--allmatches] [--nodeps] [--noscripts]
[--notriggers] [--repackage] [--test] PACKAGE_NAME
MISCELLANEOUS:
rpm {--initdb|--rebuilddb}
rpm {--addsign|--resign} PACKAGE_FILE
rpm {--querytags|--showrc}
rpm {--setperms|--setugids} PACKAGE_NAME
select-options
[PACKAGE_NAME] [-a,--all] [-f,--file FILE]
[-g,--group GROUP] {-p,--package PACKAGE_FILE]
[--fileid MD5] [--hdrid SHA1] [--pkgid MD5] [--tid TID]
[--querybynumber HDRNUM] [--triggeredby PACKAGE_NAME]
[--whatprovides CAPABILITY] [--whatrequires CAPABILITY]
query-options
[--changelog] [-c,--configfiles] [-d,--docfiles] [--dump]
[--filesbypkg] [-i,--info] [--last] [-l,--list]
[--provides] [--qf,--queryformat QUERYFMT]
[-R,--requires] [--scripts] [-s,--state]
[--triggers,--triggerscripts]
verify-options
[--nodeps] [--nofiles] [--noscripts]
[--nodigest] [--nosignature]
[--nolinkto] [--nomd5] [--nosize] [--nouser]
[--nogroup] [--nomtime] [--nomode] [--nordev]
install-options
[--aid] [--allfiles] [--badreloc] [--excludepath OLDPATH]
[--excludedocs] [--force] [-h,--hash]
[--ignoresize] [--ignorearch] [--ignoreos]
[--includedocs] [--justdb] [--nodeps]
[--nodigest] [--nosignature] [--nosuggest]
[--noorder] [--noscripts] [--notriggers]
[--oldpackage] [--percent] [--prefix NEWPATH]
[--relocate OLDPATH=NEWPATH]
[--repackage] [--replacefiles] [--replacepkgs]
[--test]
Each of the above options is explained in detail, but there are no FAQ-style examples. It's no wonder the distributions are putting layers on top of rpm (urpmi, yum, etc). I like yum, and largely because the man page is short and to the point! It only takes a few seconds to figure out what I need to do.
bp
Well, that would be better:
for image in *.tif; do convert $image -border 50 -bordercolor \#FFFFFF -quality 100 -scale 25% -resolution 96 ${image%.tif}.jpg; done
no need to do `ls *.tif`, and the bash specific construct ${image%.tif} removes a trailing occurence of the word '.tif' from the string $image