Re: message-confirm-send

[Russ Allbery <rra@stanford.edu>]

jidanni@jidanni.org writes:

> Don't tell me all you manly man types run around all day without at
> least one of
> $ alias
> alias cp='cp -i'
> alias mv='mv -i'
> alias rm='rm -i'
> $ set -o
> noclobber       on

The most useful analysis I've read of this sort of UI feature is that the
usefulness of confirmation is inversely proportional to two things:

* The frequency of the action.  "Do you really want to delete your entire
  authentication database?" is a more useful question than "do you really
  want to list the contents of this directory?"

* The ability of the confirmation logic to recognize common cases and
  *not* ask for confirmation for them, only in cases that are unusual and
  for which there's some reason to be unsure.

The usefulness is *not* as clearly correlated with the severity of the
operation.  Frequency is far more important.  This is because confirmation
of frequent operations becomes automatic.  Adding confirmation before
sending a message will, for most people, essentially just change the
keystroke for sending a message from C-c C-c to C-c C-c y.  One becomes
trained to enter the whole sequence each time, and then one ends up
confirming even when one didn't want to.

I've seen this behavior all the time with things like rm -i.  People get
so used to always entering y that the confirmation is useless for
preventing mistakes.

So, to look at those aliases again and judge them against the above
criteria, I consider the rm -i alias to be useless.  It just prompts for
every file deletion and trains one to type y automatically.  On the other
hand, the cp -i alias and the noclobber setting I use personally, since
overwriting a file is *not* the common case for cp, or for redirects, and
while I become trained to enter y automatically after cp when I know I'm
overwriting a file, the confirmation successfully catches cases where I
didn't realize I was overwriting a file.

By that criteria, confirmation on C-c C-c, since it happens every time
without any recognition of common cases and attaches to a frequent
operation, is useless in much the same way that the rm -i alias is
useless.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>