Re: [gentoo-dev] rm tragedies: a modest suggestion

[Todd Berman]

So when you have been up for 30 hours you are going to remember to type 
-c? cause aliasing will break those same shell scripts...

this is a pointless debate in my opinion, especially one to have here 
over a course of days. If you want a safe rm, alias -i and accept 
breakage, type -i, or just dont use rm...

--Todd

purslow@×××××××××.ca wrote:
> 040530 Thorsten Kampe wrote:
>
>>* Christian Gut (2004-05-30 11:40 +0100)
>>
>>>oh well, and it should move everything to a recycle bin
>>>and asking two times if you really, really want to delete something.
>>>Hey, this is not windows.
>>
>>It has to do with security.
>>Command line deletion is always more dangerous than GUI deletion
>>because you cannot see and physically mark the files you want to delete.
>>The problem is simply that computer and human beings behave differently
>>when doing multiple things at the same time or for a long time.
>>Computers don't care how long they've been uptime,
>>how many things they do simultaneously
>>or if someone is disrupting their concentration.
>>People do - and there is nothing to eliminate these things;
>>just to put a barrier to the possible disastrous consequences.
>
>
> these msgs seem to sum up the opposite sides of the debate.
>
> the point i wanted to raise has nothing to do w novices, grannies or M$ :
> it's something which can affect the most experienced red-blooded sysadmin
> simply because for once in his life he's been working too long
> & just before he collapses he enters a cmd w an extra space in it.
> when machine guards, safety hats/boots etc were introduced in factories,
> there were similar objections from old hands who "didn't need such things",
> at least till they ended up in hospital missing a foot/hand/eye.
>
> yes, there are good reasons not to alter the basic default behaviour of 'rm':
> eg thousands of scripts exist out there which depend on it.
> however, that doesn't mean that additional flags can't be added to it.
>
> what is needed is a flag whose action lies between '-f' & '-i':
> let's call it '-c' (for 'check': there are lots of spare letters to use).
> 'rm -f' & 'rm -i' would continue to do just what they always have done,
> but unlike '-i', which asks re each file/dir individually
> -- which is why people don't want to use it for big deletions -- ,
> '-c' would ask just once for the whole set:
> it would show the list (after expansion by the shell) & the starting dir,
> with a request for confirmation like that offered by Zsh; for extra safety,
> the user would have to type 'yes', not just 'y', to confirm deletion.
> to allow easy aliasing of 'rm -c' as 'rm', 'rm -c' wd ask for confirmation
> just when multiple files were listed or the list was a directory
> -- esp this would apply when '*' was expanded by the shell -- ,
> but not when there was only  1  matching file (not a dir).
>
> this sb enough to alert the guy who's been working  30 hr ,
> but would have no effect on all those thousands of scripts,
> which wd continue to work with 'rm -f' just like now,
> & no-one wb forced to use '-c', if they were willing to take the risk.
>
> "But that's not POSIX !" -- nor is the GNU 'rm' now used by Gentoo,
> which has a couple of extra flags which POSIX doesn't know about.
>
> so is there any reason -- apart from developer time -- this can't be done ?
>


--
gentoo-dev@g.o mailing list