Bugzilla – Bug 59038
sax2: Ctrl-Alt-Backspace to kill X should be disabled by default
Last modified: 2007-09-07 08:05:13 UTC
When we generate the XFree86 config, we should set Option "DontZap" "yes" In the ServerFlags section of the config file. This disallows the use of the Ctrl+Alt+Backspace sequence to kill X. This is something most users do not know about anyway, and if one wants it, it can be turned back on. Especially when we have virtual desktop keybindings use Ctrl-Alt, it is not very hard to accidentally for example Ctrl-alt-cursor to your word processor, then start erasing your text with backspace just a moment before you release ctrl-alt.. *poof*.. This is probably an Yast issue for the configurator, but I am just learning my way around the suse bugzilla, so I apologize, please help me to get this filed to the correct component :)
I'm now maintaining our X11 packages since about 5 years and I can't remember that any user ever complained about this. To be honest, I still can't believe this story. It's a feature that often is useful to be able to kill your Xserver. Therefore we won't change this.
Consider me the first one to complain. It has happened to me a bunch of times, often enough to be annoying. Wonder why I decided to file this today? :) But uh, if you dont even know you can kill your X with the key combo (which most users do not know about) - it is not useful at all. It's a completely hidden feature. Thus it should be disabled by default, those who need it often (like people who do hacking and get X pointer grabs with crashing code a lot, they can enable it. Normal users do not encounter the need to kill their X all that often really. I think it should be disabled. It's not like the feature will be removed completely.
Just because it is hidden doesn't mean it isn't useful :) But yes, in general, I agree; others have complained as well- it does make it too easy to accidentally nuke your server. Tuomas is not the first I've seen do this, though admittedly probably the first 6-7 year user of Linux I've seen do it :)
For the most part I agree with Tuomas randomly destructive keystrokes that are under-documented and do not give confirmation probably shouldn't be included in a professional product. (No! Don't press the red button! It launches the nuclear missiles! For a better example of humor here, see: http://www.ok-cancel.com/strips/OKCancel20040709.gif) Anyway, in the end I think we should allow C-A-BKSP to exist anyway. It serves as a security feature of sorts. You can always press that key combination to get back to the REAL login screen. In this way you can always ensure that you aren't logging in with a trojan.
There is another reason not to change this. For sure we won't change this for SLES, but this means that we would need to maintain different XFree86 and/or SaX2 packages for SLES9 and NLD. Neither Marcus (SaX2) nor me and our project manager want this I'm pretty sure.
Why would this feature *need* to be present in one product and not in another? Ben: as a security feature it is kind of silly, as if you hax0red a box, you could install a hax0red login manager which just starts again even if you restarted X. I am just fearing that the ctrl-alt part of desktop switching keybindings DOES make it a bit too easy to accidentally press backspace while you happen to be holding those two keys already down. Geeks do not complain about this - they just disable it. My point is: what if your car had a feature, so that when you pressed the clutch pedal, switched to 5th gear and simultaneously opened the glove compartment while tuning the radio, the engine would shut down and the steering lock would engage? Sure, it does not happen much, and frankly I dont think there would be many people left complaining if it happened to them.. I mean, the feature is hidden from most people anyway, those who need it know how to enable it. Why should it be turned on by default? The answer "it's always been like that" is not the point.
*** Bug 305999 has been marked as a duplicate of this bug. ***
Focus of discussion for some opensuse people now, moving the bug to opensuse 10.3 as I don't see much private in here. Comment #5 can be reshaped as SLE vs openSUSE instead of SLES vs NLD9 now.
I find the user pressing by accident argument somewhat more compelling than the locked workstation argument, as there is no warning at all before X is killed. In any case perhaps it could be configurable somewhere like where the sysrq on/off toggle is. Interestingly we default sysrq to disabled, even though they are also useful for killing things/rebooting. http://benjiweber.co.uk/yastSecuritySettings2.png is where sysrq can currently be turned on and off and http://benjiweber.co.uk/yastSecuritySettings.png could have one of the profiles with it off, and others with it on. I think if this is disabled then there at least needs to be a relatively easy way to turn it back on because a lot of people use it.
I'm reopening this for openSUSE 10.3. Will you please reconsider adding the DontZap option as default? Yes, the Ctrl+Alt+Backspace "feature" is very useful to developers/hackers/power-users, but the real problem comes when people discover it accidentally. On my keyboard (the Ctrl and Alt keys are right next to each other) I've hit this problem way too many times. I'll be working along and suddenly, poof, my session is gone! There is a much safer way to kill a locked X server. Just Ctrl+Alt+Function Key to access a terminal, log in as root, and run the kill command. While this isn't as convenient, it still allows a user to kill X and doesn't affect a standard non-poweruser. Bug 305999 is another case where this could be potentially hazardous to a user's data. A more believable use case is when my 2-year-old son rushes to my keyboard when I step out of the room for a couple seconds, presses a bunch of keys at once and happens across Ctrl+Alt+Backspace. Poof! Just lost my session/work. And yes, the Ctrl+Alt+Backspace is currently in full effect when my screen is locked with a screensaver.
I'm against this change, but I decided to no longer waste my time on useless discussions. If anyone wants to comment feel free to do so. Our project manager needs to take the decision. --> NEEDINFO
I'm against this change too. I don't see why people would accidentally hit Ctrl-Alt-BackSpace and most of the Linux users I know want to have the ability to kill the server by this control keys
(In reply to comment #12 from Marcus Schaefer) > I don't see why people would accidentally hit Ctrl-Alt-BackSpace Like I mentioned above, my Ctrl and Alt keys on my keyboard are right next to each other. One of the keystrokes I use frequently is Ctrl-Backspace. It deletes the previous word in most word processors, web browsers, etc. If I'm clipping along at a really good speed and my left pinky finger happens to not go too far left, it's all over and my work is lost. I can't count the number of times that's happened to me.
>Interestingly we default sysrq to disabled, even though they are also useful >for killing things/rebooting. I would love to see this enabled by default. Always when you need this feature you need to recognize that it is not the default. But this is another discussion I gave up a long time ago. At least nowadays we have journaling filesystems ...
I find this discussion pointless. If you do a ctrl-alt-del in text mode the system reboots. There is nothing to stop it to do this. If you did the same on some earlier Windows version (and in the old dos days) the same thing happened. Yes, one could do something to pop up a window and let the user confirm. This would involve some work - but it would be feasable. Yes, I have disabled this on my main workstation, too as I often test servers which i need to zap and accidentally do this on the wrong keyboard or with the kvm switch switched to the wrong keyboard. However I keep this enabled on all other machines, especially at my workstation at home as I never have had the problem of accidentally hitting this sequence. I often find myself using this sequence when some applications have hogged the Xserver so badly that there is no other way of getting rid of it (and the entire session). I know numerous people who use this as the standard way of terminating an X session instead of going thru a lenthgy logout procedure. Therefore I assume that we will see a lot of bug reports if we disable this feature by default. So I opt against this.
(In reply to comment #15 from Egbert Eich) > I find this discussion pointless. If you do a ctrl-alt-del in text mode the > system reboots. There is nothing to stop it to do this. If you did the same on > some earlier Windows version (and in the old dos days) the same thing happened. > > Yes, one could do something to pop up a window and let the user confirm. This > would involve some work - but it would be feasable. Great suggestion! A pop up window would be excellent. I assume if X was really locked up it wouldn't be possible to pop open a window, in which case, perhaps multiple presses of Ctrl-Alt-Backspace would then cause a crashed X to really die? But, for normal operation, a pop up could at least prevent most of the accidental problems.
I should point out that it's probably too late to make this either easily configurable or have any sort of popup for 10.3 as suggested due to string freeze.
Talking with Stefan, Egbert and Christoph, we agreed on the following: Marcus, please change the default to disable this. We'll mention this change in the release notes (see bug #306265). The discussion here was rather controverse, so we will make this change for beta3 - and consider to reevaluate this after beta3 depending on the reaction we get for changing this long time default.
*** Bug 306265 has been marked as a duplicate of this bug. ***
Note: This happens against my wish and I will not take over any responsibility for this change ! Additionally I have to admit that I don't like that the opinions of developers are not respected here !
Do we have any mechanism for users/administrators adjusting this, other than editing xorg.conf?
(In reply to comment #21 from Benjamin Weber) > Do we have any mechanism for users/administrators adjusting this, other than > editing xorg.conf? No. At least not for openSUSE 10.3. comment #15 > [...] > Yes, one could do something to pop up a window and let the user confirm. This > would involve some work - but it would be feasable. This would be an option for a future product.
> and consider to reevaluate this after beta3 depending on the reaction > we get for changing this long time default. We can influence this. :-)
Thanks for implementing this feature. Once in a while it also happens to me that I accidentially kill X session by pressing this key combo.
This seems familiar like Ubuntus's BulletProofX ( https://wiki.ubuntu.com/BulletProofX ) which is currently discussed on opensuse-factory. I would like to suggest to give the user the choice whether he/she needs this feature or not. Therefore it could be proviced as an advanced option in the installation process. Someting like this: *------------------------------------------------------ | [ ] Terminate X-Server via Alt-Ctrl-Backspace | [ ] SysRequest enabled by default | ... and other options which could be asked here... | (If you aren't sure what this is, continue anyway) *------------------------------------------------------ So only if the user/developer needed this, he/she would enable it.
I disagree with the simple disabling of this key combo. The clueless n00b writing the term paper can always break things with other key combos - will he know how to get back to the desktop from an accidental ctrl-alt-f1 - will he guess ctrl-alt-f7 before trying ctrl-alt-delete? Also, I'm sure I've hit alt-f4 and then Discard by mistake in a hurry before. And marauding 2 year olds and computers just don't mix - beakers of orange juice and pushing laptops off the desk spring to mind. If a graphical confirmation dialog to prevent the accidental case could be implemented later, I'm for that, but just disabling ctrl-alt-delete is insufficient for the stated purpose and prevents legitimate uses.
(In reply to comment #27 from Will Stephenson) > If a graphical confirmation dialog to prevent the accidental case could be > implemented later, I'm for that, but just disabling ctrl-alt-delete is > insufficient for the stated purpose and prevents legitimate uses. I agree. If the intention of disabling it is to see how many power users we can get upset, we're going about it wrong. We don't need a bunch of bug reports from power users just to prove the feature is useful. I think all the comments here have shown that there are many who actively use the feature. Let's figure out a way to make both the naive and power users happy at the same time. It's too late to solve this the right way for 10.3 because of the feature freeze, so perhaps we should evaluate this for the next release.
That's unrelated, Will. I know about ctrl-alt-Backspace, but once or twice the year it happens to me that I press this key combo without intention. It happens because on some laptops ctrl and alt are close to each other, and on my workstation keyboard they are not. If you "discard by mistake in a hurry," that's something different. Anyway, it is all in this thread... If you think displaying a confirmation dialog will be possible, fine with me. Otherwise consider to move the ctrl-alt-Backspace feature to a 4-key combo (I do not know whether that's possible, though).
I must join the protesters. First of all I don't agree with the premise that noone knows about the feature - any noob is told about it first time he tries to install some 3D-driver or doing other xorg-configuration or when desktop effects freeze his X or something similar. It's very frequently used. Secondly I've never heard about nor seen the problem ever, in 2-3 years of being a front line soldier on IRC, webforums, usenet and LUG meetings. It seems only Novell employees with strange laptops without "Windows"-keys and using power-user/hacker word processing shortcuts, that aunt Tillie never even imagined, have the problem. Furthermore other distributions, even the almighty, super cool and friendly Ubuntu, don't have a problem with having this feature enabled. So maybe the leet Novell hackers in question should just disable the option themselves manually if they have a problem, instead of dumping down openSUSE by removing/hiding functionality. Thirdly, the whole process of this is very troubling. Some of us eat, sleep, shit and talk openSUSE, and we're beginning to have rather well working project, with good cooperation and communication between employees and community members - and suddenly some people show up, whom I've never seen a single mailinglist post from or a comment in an IRC status meeting, nothing, and these people seemingly dictate what shall be done for openSUSE against the will of the package maintainers and active users. Alarming stuff.
One proposed alternative is a pop-up with a timeout: "X will restart in 10... 9... 8..." [OK] [Cancel]. That both preserves the current behavior (including the case when the keyboard/mouse are unresponsive, with a minor added inconvenience) and prevents typos from immediately killing X.
FWIW, there's a good discussion about this subject here: https://wiki.ubuntu.com/XorgCtrlAltBackspace Although it's more of the same discussion, it does have documented cases of accidental discovery.
(In reply to comment #32 from Mark Gordon) > One proposed alternative is a pop-up with a timeout: "X will restart in 10... > 9... 8..." [OK] [Cancel]. That both preserves the current behavior (including > the case when the keyboard/mouse are unresponsive, with a minor added > inconvenience) and prevents typos from immediately killing X. This sounds good, but I agree with #27 - as it's too late to do it properly for 10.3 (string/feature freeze) why not postpone the change to 10.3++ ?
Ok, let's do it properly after 10.3. I'm filling a proper enhancement request for this now.
FATE #302543.
Closing.
Closing before the change/submission is reverted?
sorry for being to pushy here.
(In reply to comment #16 from Boyd Timothy) > > Yes, one could do something to pop up a window and let the user confirm. This > > would involve some work - but it would be feasable. > > Great suggestion! A pop up window would be excellent. I assume if X was No, it wouldn't. As Egbert said many (many!) people use this as a shortcut logout procedure. A popup window complete contradicts this. Normally people don't have any problems at all with this combination, and those who have can always add NoZap. Actually, this key combination should be configurable, but at the moment it isn't (because in the beginning nobody upstream thought it would ever be a problem, later nobody cared). Also moving it to a different combo might be problematic with power users (who never read release notes ;) , but might be the best solution. E.g. adding shift to the modifiers. Not all keys can be detected correctly for more than two keys pressed in a typical keyboard matrix layout (most keyboards use that), but the modifier keys are typically not checked by the matrix but individually.
(In reply to comment #40 from Matthias Hopf) > (In reply to comment #16 from Boyd Timothy) > > > Yes, one could do something to pop up a window and let the user confirm. This > > > would involve some work - but it would be feasable. > > > > Great suggestion! A pop up window would be excellent. I assume if X was > > No, it wouldn't. As Egbert said many (many!) people use this as a shortcut > logout procedure. A popup window complete contradicts this. > > Normally people don't have any problems at all with this combination, and those > who have can always add NoZap. A warning popup dialogue could have a "don't show this again" option, which would make it discoverably configurable.
> ------------------------------------------------------------------- > Thu Aug 30 10:36:20 CEST 2007 - ms@suse.de > > - [...] > - added option DontZap [...] (#59038) Marcus, I would like to ask you to revert this change in sax2 for Beta4. The bug can be closed as FIXED aferwards. Anything else needs to be handled via FATE #302543.
done