Fwd: Is disabling ctrl-alt-backspace really such a good idea?

[John Moser]

On Tue, Feb 10, 2009 at 10:57 AM, Justin M. Wray
<wray.justin.ubuntu at gmail.com> wrote:
> Even if a new user is unfamiliar with the key combination, it only takes a little education OR them doing it once. Lessons are learned hard.
>

This is poorly conceived thinking, but switching it off is straight up
naive.  "This is a power user tool and can accidently happen, and is
bad when accidents happen, so let's take it away."  Knee-jerk reaction
to dumb down the system.

The correct solution is not dead-simple to implement, but it works:
When you C-A-B, grab the whole screen and put up a confirmation dialog
like gksudo does.  You can't switch desktops off it (CAB happens
sometimes while swapping around virtual desktops and landing in a text
editor, trying to backspace over stuff), and it's right in your face.
It says "In 10 seconds I will kill X and reload it.  This will destroy
all of your work.  Click the big red CANCEL button below now or hit
SPACE!"

If the X server is frozen, the dialog can't be shown, can't be
clicked, can't be touched.  All you have to do is modify X to take an
option "ZapProgram" "/usr/bin/uzapme" which gets the $DISPLAY
environment passed to it.  uzapme reads the user's configuration,
executes a child "gzapme" or "kazzapme" and waits 10 seconds for the
child to return.  If the child doesn't exit favorably, i.e. complains
it can't draw on X or just hangs (because of no user input, or because
it tried to talk to X and got hung up), uzapme kills it and returns an
unfavorable state to X.  X then executes the ZAP code.

This works because it relies on X's zap code to be executed.  If CAB
can cause X to exit, then it can instead cause X to run a child
process and wait() for a result.  If doing such would inherently hang
and fail to terminate X under conditions of the child returning status
"ESCREWTHISTERMINATEX" then obviously X is so borked right now that
the zap code wouldn't have functioned anyway and you would have had to
unplug it and plug it back in.

> Should we remove all the abilities that may "damage" the system?
>

No, just throw up a single safety net.  For many unsafe operations
this is '-f', so on a real (not Fedora) system you 'rm -r /' and it
goes "do you want to remove these files?" because you probably did
something damaging; whereas you 'rm -rf /' and it goes "oh, you know
what you doing, move zig."  If the user's dumb enough to use
'--force-all' or not click a button marked "Press this or your
computer is effectively going to crash," it's now officially his
fault.

> Where does the line get drawn?
>

Usually in the sand, but sometimes on paper.  In HTML the line gets
drawn at an <hr> tag.

> The C-A-B is an easy thing to learn and avoid, but a powerful resource when needed.
>

Same argument for magic sysrq keys... not for non-power-users, some
people just want the OS to get out of the way.

The REAL solution is to run everything through an X proxy that can
lose the X server head, like 'screen' for X.
> Sent via BlackBerry by AT&T
>
> -----Original Message-----
> From: Dylan McCall <dylanmccall at gmail.com>
>
> Date: Tue, 10 Feb 2009 07:02:51
> To: Clive Wagenaar<clivewagenaar at gmail.com>
> Cc: <ubuntu-devel-discuss at lists.ubuntu.com>
> Subject: Re: Fwd: Is disabling ctrl-alt-backspace really such a good idea?
>
>
>> I agree with this guy, have it on by default, noobs can use GUI to switch it
>> off.
>>
>> Else, millions of users will be doing this on first boot up:
>>
>> alt f2
>> gksudo gedit /etc/X11/xorg.conf
>>
>> Section "ServerFlags"
>> Option "DontZap" "no"
>> EndSection
>
> The important thing is that those millions of users actually don't
> mind tinkering with xorg.conf and probably do anyway. The users we are
> trying to help, however, Don't Know The Key Combo Exists, or that
> xorg.conf exists, or that they need to explicitly tell the system how
> to be easier to them, until it is too late. They don't want to do any
> extra configuration after installing the operating system. Those
> millions of power-users do.
>
> It astounds me how people just forget about Ubuntu's goals and
> aspiritions in light of this issue. It isn't doing much for user
> friendliness when the community of contributors is using bad names on
> those new users Ubuntu strives to be gentle to and then treating them
> like outsiders.
>
> It isn't /likely/ for someone to hit C-A-B (although it's been done
> once or twice by yours truly, particularly with graphics tools), but
> the immediate issue is that we have a key combination which, without a
> moment's question, eradicates one's session from existence. No session
> saving, no notice, no sensitivity to whether the keyboard is grabbed
> or another application is handling the event. It just happens no
> matter what. Further, it relies on common keys. Sysrq K is alright
> since nobody tends to use the Sys RQ key, but Ctrl and Alt are both
> everyday modifier keys and Backspace is a natural key for deleting
> stuff. We want our users to feel free exploring Ubuntu without the
> risk of wiping out the system (within reason, of course). Hopefully
> they can gain a trust of themselves and the system that way, start
> paying more attention to the text on the screen and learning what it
> all means. One thing I know is that a single catastrophic event like
> "tinkering led to the loss of two hour's work when I pressed Ctrl Alt
> Backspace" causes someone to doubt the value of that exploring. Then
> there's another user reliant on others for resolving all issues
> related to the computer.
>
> Something interesting I've learned in an Ubuntu Forums thread is that
> a surprising number of people who want this key combo to stay don't
> actually use it for its intended purpose (to reset X when it is
> crashed). Instead, they use it as a shortcut for logging out. Doing
> that is risky, messy and inadequate.
> Perhaps if logging out (with session saving) was mapped to Ctrl Alt
> Backspace we wouldn't have as many bothered users.
>
>
> Bye,
> -Dylan
>
> --
> Ubuntu-devel-discuss mailing list
> Ubuntu-devel-discuss at lists.ubuntu.com
> Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
> --
> Ubuntu-devel-discuss mailing list
> Ubuntu-devel-discuss at lists.ubuntu.com
> Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel-discuss
>