`alias rm="rm -i"` considered harmful?

Question

I have read some time ago (can't find the reference) that using such an alias as alias rm="rm -i" was very bad.

Is there historical evidence or common sense explanation for that fact?

I would imagine that it gives a user a bad habit of relying on the confirmation prompt to check his command, which could lead to disasters if he does so on another profile that doesn't have the alias.

Answer 1

You're right.

It's bad because you get used to it. If you're on a system that doesn't have it, and you rm, it immediately starts deleting and you're wondering what's going on.

Many users are used to SSH'ing into different systems; so using lots of different systems, sometimes without personalized user accounts (including aliases) set up, is rather common.

Instead, use e.g. alias rmi='rm -i' and learn to use that one. If that isn't set up on a different system, you didn't accidentally delete files and can always fall back to typing the full command.

Answer 2

Like @Daniel said, it's not harmful in and of itself, other than training you to expect it to be there. In fact, it's the default set-up on CentOS (and by extension RHEL, I presume - been too long since I've used one) machines, and it's a massive pain in the tuchus. For the rest of my time at that gig, I typed /bin/rm in order to avoid the "Linux for people who shouldn't have root access" setup.

Answer 3

I think the big danger is that people might rely on something like this to filter a glob. Imagine you want to delete some images from a directory, but not all of them:

rm -i pics/*.jpg

You could use that to filter the glob manually, which would be completely reasonable. But, if you had aliased it and were using rm and happened to land in a shell without that alias and try it... you just deleted all your pictures, oops!

Personally I also find this alias to be harmful to my blood pressure ;). But that's just me.

Answer 4

In addition to what Daniel Beck said, I found myself using -f to bypass the -i, which is potentially dangerous as it results in using rm -f and rm -rf unnecessarily. Somehow related: a way to prevent rm -rf issues is this is to create file called "-i" as duscissed in answer of this question:How do I prevent accidental rm -rf /*?.

But again, if that alias wasn't there, I wouldn't use -f, and the whole thing won't be an issue.

Answer 5

This is much less harmful, based on my experience working with hundreds of users in the past:

rm ()  # must be a function, must require single answer for all targets
{
    ls -FCsd "$@"
    local reply ; echo -n 'remove[ny]? ' ; read reply
    if [ "_$reply" = "_y" ] ; then
        /bin/rm -rf "$@" ; else echo '(cancelled)'
    fi
}

• Users are trained to use wildcards correctly, not just '*' and then relying on the y/n prompts to select files
• The conditioning to use correct wildcards often saved the from disaster when they used rm in some other context that lacked either this function or the rm -i alias.
• I spent less time restoring files where the user typed 'y' one too many times
• Users only have to respond once - providing a sharp positive feedback to their using it
• Control-c interrupts work and are reports as doing nothing
• Not a script, so the real rm is untouched, leaving other programs undaunted.

The code style is mostly sh-compatible (except use echo .... | tr -d '\012' for pre-bash shells), feel free to make your own more bash-specific. I'm not posting to share the code itself, but to share the user experience change that comes with it.