Death of the Command Line

    August 1, 2006

It’s hard for me to imagine using an OS without a strong command line. Even Microsoft has recognized the for that with their Monad Shell ( though they are at least temporarily removing that from Vista).

Linux of course has its Bash shell, Mac OS X has Terminal (which now defaults to Bash) – everybody knows you need a shell.

But who really uses it? Well, I do, and maybe you do, but most Windows users certainly don’t, and most Mac users are equally command line phobic. Linux users may be a bit more apt to drop into typing-land, but even there its being pushed into insignificance.

For example, I recently installed a Suse 10.1 system. I needed to add some printers, but I couldn’t remember the syntax for adding an smb printer with lpadmin – it’s just “lpadmin -p printer -v smb://host/printer” but I was having a senior moment. I was sshed in at a command line and what I wanted was something like “redhat-config-printer” to help me through it. If there is such a thing in Suse, I couldn’t find it.

I was able to get by this by installing Lynx (yast2 –install lynx) and running “lynx http://localhost:631” to get the Cups browser interface (though I first had to add a cups user with “lppasswd -a root”). That’s fine; I was able to do what I wanted, but I’m pretty sure most people would have given up and used the GUI.

We old command line hands know all the arguments for the importance of the command line. We can build tools on demand with pipes. Therefore we can do exactly what we want quickly and efficiently. GUI tools often have less features and can’t be piped to other GUI tools.

Yeah, but.. They could: Nothing stops a GUI tool from implementing all the features of its command line equivalent. And there’s no reason that a GUI application couldn’t pass or accept data from some other GUI application. If all applications were written with that ability, you could just string things together. In fact, some graphical programming environments work just like that: you drag around little tools and create bigger tools without suffering the indignities of that awful command line.

If every GUI app could read standard input (and actually many do) and had an option to send text output to any other tool, we’d have the same power in the GUI. You’d need a master pipe fitting app to build reususable tools, and of course the tools themselves would need to accept behaviour changing arguments, but none of that is difficult.

Would this be better than the command line? Worse?

Added after the original post to avoid more misunderstanding:

I’m not a “point and click” fan. I LIKE CLI’s for their power, speed and convenience. My fear is that developers are neglecting CLI tools in favor of GUI versions.

*Originally published at

A.P. Lawrence provides SCO Unix and Linux consulting services