RogerBW's Blog

Many modern Linux systems assume that you will never have a root shell. Instead, you are expected to prepend "sudo" to every root-type command.

This is a safety measure, and I can appreciate that. But it seems to give up a great deal of capability (for example, the ability to scan down a file hierarchy and tab-complete a long boring device name rather than typing it out in full) in return for that.

The SR-71 pilots used to say that if you weren't just a little bit scared getting into the cockpit, it was time to retire. That's the way I feel about a root shell: it's scary, and it should be. You can destroy the entire machine with a single command. But that's why you're running Unix in the first place: a tungsten-carbide disc grinder will cut through more things than a pair of plastic safety scissors, and those things need not include your hand.

I suppose sudo is fine for copying and pasting someone else's recipe (though if you're doing that you're effectively trusting them with your machine anyway), or for doing a set of standard configuration commands when you're setting up a particular bit of software. But for troubleshooting, where you may well need to have multiple rooty processes running and talking to each other, it feels like just a waste of keystrokes.

(The modern sysadmin approach of rebuilding the machine image from scratch, rather than trying to work out why it went wrong and fix it… is a subject for a separate rant.)

If you're running your own system rather than following someone else's instructions, you'll simply get into the habit of prepending sudo to certain commands. There will be nothing unusual about it, no frisson to remind you that you're playing with the grown-ups now. Make fear work for you: the # prompt can be a useful red flag.