69981 - backspace/delete is mapped by default to go back

Status: RESOLVED WONTFIX

(Core :: DOM: UI Events & Focus Handling, defect)

Description

[Ian Tommins ('thelem')]

When a user presses the backspace key, they are sent back a page in the history.
This matches IE's shortcut mapping. I think this is a bad shortcut and should be
removed because:
 * A sutable alternative exists (alt + back arrow. That should probably be CTRL
instead, but that is another bug).
 * There is no obvious partner for the forward button, in the way the left and
right arrows are an obvious pair.
 * This makes the backspace key do different things in different places. eg the
user loads a page containing a text box, clicks in the first box to start
typing, but changes their mind and presses delete to got back. Nothing happens.
Anyone who had used IE for an extended period will know this can bve very
annoying. In my case, the problem is that I click in a form, press backspace and
it goes back a page.

I'm sure that there are other arguments, but I can't think of them now.

Advantages for keeping this action are:
 * Parity with IE (should be a minor point, that sways something which people
are not sure about)
 * A single key shortcut (convenient, but causes bugs mentioned above).

Comment 1

[Ian Tommins ('thelem')]

Adding CCs requested by mpt

Comment 2

[Alec Flett]

over to aaron lev. for key issues. Bugzilla is not a place for design issues
anyway, not sure this is actually a bug.

Comment 3

[Claudius Gayle]

1. this is a session history thing not a global history thing.
2. we have a whole component called UI: Design feedback.
3. I agree with the reporter. It's overkill and doesn't make sense without the forward cognate.

Comment 4

[Boris Zbarsky [:bzbarsky]]

4. looks like a duplicate of bug 60102

Comment 5

[Jesse Ruderman]

See also bug 36922, "Delete/Backspace to scroll a browser window up".  I'd 
actually rather that space/backspace didn't do anything at all, since there are 
already pgup/pgdn keys, and those can work even when textfields are focused 
(and also happen to be close to each other and the arrow keys).

Comment 6

[timeless]

i still like space. what we really  need is to allow users to change bindings.  
I have an idea about how to change menus, and I think it could probably be 
expanded to allow users to change bindings...

Comment 7

[Aaron Leventhal]

Keyboard UI discussion is going on over on the mozilla-accessibility mailing
list. We'll have a newsgroup soon.
send a message to [e-mail]  subject = subscribe

I kind of like Space and backspace the way they are, personally. My 2 cents.

Oh yeah, take a look at the spec:
www.mozilla.org/projects/ui/accessibility/mozkeyintro.html

Comment 8

[Matthew T (active 1999-2002)]

Where a key (Backspace) was designed specifically for deletion, making it do 
anything else -- ever -- is messing with people's heads.

--> UI:DF

Comment 9

[Aaron Leventhal]

I don't agree. If someone else wants to take this bug, they can reopen it.

Comment 10

[Zach Lipton [:zach]]

I kindof have to affree with mpt. Verif

Comment 11

[Matthew T (active 1999-2002)]

> If someone else wants to take this bug, they can reopen it.

Ok then, I'll take it, since nobody here has agreed with Aaron on this one. 
Imitating Internet Explorer is good for learnability, but we don't need to go 
so far as to repeat their mistakes.

Comment 12

[Matthew T (active 1999-2002)]

Mine.

Comment 13

[Stephen Ostermiller]

See bug 75972 for problems with this UI in relation to the flash plugin.

Comment 14

[Gervase Markham [:gerv]]

mpt: This is a remove-2-lines patch, in:
http://lxr.mozilla.org/seamonkey/source/xpfe/browser/resources/content/win/platformNavigationBindings.xul

If you want me to make it up, shout :-)

Gerv

Comment 15

[Ben Goodger (use ben at mozilla dot org for email)]

Don't do this.

Comment 16

[Ian Tommins ('thelem')]

Why not Ben? It also solves the delete in flash forms bug.

Comment 17

[Matthew T (active 1999-2002)]

One day last week, a customer was using Hotmail, and had written rather a long 
message, but got an error when he tried to send it. The error was that 
`john.smithahotmail.com', which he was trying to send to, was not a valid
e-mail address. (It wasn't exactly that address, but you get the idea.)

He asked for help, and I pointed out that the `a' which he'd copied off a piece 
of paper was actually supposed to be a `@'. So he clicked to the right of the 
`a', and pressed the Backspace key.

But for some reason, the mouse click hadn't focused the text field. Possibly 
this guy was a couple of pixels off the field when he clicked, I'm not sure 
exactly. Anyway, the result was that focus was still on the page as a whole. So 
what did Backspace do? Instead of deleting the `a' character, it took him from 
the message composition page right back to his Inbox. The message which we'd 
both been looking at a moment earlier had seemingly vanished completely.

Both he and I were rather confused by this, but after a few seconds he worked 
out what he needed to do -- he needed to get back to the composition page, so 
he could try to send his message again. He was perilously close to clicking the 
`Compose' button to do this, thereby removing the previous instance of the 
composition page from the cache and wiping his message completely, when I 
noticed that the `Forward' button on the toolbar had become enabled. I realized 
what had happened, and stopped him just in time. A mouseclick on the `Forward' 
button later, and we were back at the composition page. Tragedy had been 
averted for another day.

I discussed this with Aaron on IRC. He agreed that Backspace should not go Back 
by default, while saying that he would still have Backspace set up to go Back 
in his own personal set of keybindings. Which is fine by me.

Reassigning to Gerv. Note that this does *not* solve the delete in Flash forms 
bug, it just gets rid of that bug's most irritating symptom.

Comment 18

[Gervase Markham [:gerv]]

Attachment: Patch to remove Backspace and Shift-Backspace bindings on Windows

Comment 19

[Gervase Markham [:gerv]]

Patch attached. Note: because my build environment is Unix, this patch has not
been tested. However, if this is not the correct patch, then it darn well should be!

Gerv

Comment 20

[Gervase Markham [:gerv]]

mpt: are we still trying to do this?

Gerv

Comment 21

[Matthew T (active 1999-2002)]

Don't see why not. CCing Jag for r=.

Comment 22

[jag (Peter Annema)]

Yep, that's the fix. r=jag

Comment 23

[Gervase Markham [:gerv]]

blake: Any chance of a 5-second rubber stamp here?

Gerv

Comment 24

[Blake Ross]

sr=blake

Comment 25

[Aaron Leventhal]

You're not even going to make it page up the way it does in 4.x?

What's up with that?

Comment 26

[Gervase Markham [:gerv]]

Checked in.

Aaron: I would guess that the answer is that having backspace do anything,
particularly something which changes the viewport, is too risky in terms of user
confusion. But if you want that binding, file an RFE. In the mean time, let's
ditch this one.

Gerv

Comment 27

[Vicente Salvador]

I don't want backspace to do anything!!!

It produces many mistakes on pages with forms!!!

Comment 28

[Aleksander Adamowski]

Any follow-up bugs?
I understand, that the IE backspace binding was a bad thing, but Alt-leftarrow
isn't any better.
First of all, it's two keys combination, which is too many for such a frequently
used command. It should be a single key.
Second, when navigating the page, you usually have one hand on the keyboard, the
other on the mouse. The keyboard hand is located at the cursor and pgup-pgdn
blocks. Thats a bit far from the ALT. It gets worse when you are using Polish
keyboard layout (which uses right ALT for typing of diacritical characters! That
leaves left ALT which is at the other side of the keyboard and you really need
two hands to press it)

So I propose a follow-up bug to decide on an alternate mapping.

INS key maybe?

Comment 29

[Vicente Salvador]

I'm sorry for the stuff on this fixed bug bud this is just to clarify.

You'r assuming that the mouse is handled with the left hand.
Normally is handled with the right hand, so Backspace, Insert, etc. enforces the
user to leave the mouse to type the key.

So, you shouldn't assume so many things and sometines is quicker to press the
back button with the mouse than typing any key. I thing that many of the menu
keyboard shortcuts are Control+key (Compatible in all platforms: linux, win,
mac...) and ALT+Key is not so bad.

Maybe we should add another shortcut for "back page", but not with any form
editing keys like Backspace, Insert, Home, End...

A posible Shortcut for back-forward should be "Back Page-Forward Page" keys!

Comment 30

[rxbypg001]

I personally liked the backspace mapping, and I liked it when it first was
intro'd on IE. Having a single keyboard key to go back is a fantastic idea. The
fact that it's a big key is also nice, since it's the most common action I use
in the browser. Nothing else has increased my browsing speed and easy-of-use
than that mapping.

The Alt-Left combo is painfully slow in comparison. Is another key mapping
amenable or can the option be given to leave the mapping as it was? As it is,
I'm not going to be using nightlies past this point unless this issue is
resolved. The alternative is going back to IE5 (ugh).

Comment 31

[Aleksander Adamowski]

Agreed 100%. 
Are you left- or right-handed?

Comment 32

[David De Graff]

I'm just a lowly user who has been trying to help out by reporting bugs now and
then, but I am amazed that this issue came up at all. I thought that the newly
'dead' backspace key was a bug, and I reported it as such only to be told that
it's a new feature.

The backspace key link to the back button functionality had worked just fine for
many mooons in Mozilla until about a week ago, and I along with (I suspect) many
others find that mapping to be very efficient and straightforward. IE and Opera
both use the backspace key in that fashion, and although this project does not
need to slavishly follow fashion there are some UI functions that users in
general have grown to understand, expect, and use frequently. This is one such item.

Keys do different things depending on the cursor focus in many situations,
whether in the OS, browser, or other apps. I think that most users are
accustomed to this and that the benefits of an intuitive link between the key
name/general function and the navigation use outweigh any cursor positioning
confusion.

PLEASE change it back to the way that it was up until a week or two ago, and
just let it be.

BTW, I move my mouse back and forth from right to left hand when one or the
other needs a rest, and I find the backspace key to be very useful in either case.

Comment 33

[Aleksander Adamowski]

Please take a moment to read bug 93862, I think that the only solution would be 
to implement reconfigurable key bindings im Mozilla.

Comment 34

[Gervase Markham [:gerv]]

This bug is about removing these bindings, and it's fixed. 

For those who don't like this, it should be obvious from the patch how to add 
them back to your copy of Mozilla if you like. This will do until we get 
configurable key bindings.

Gerv

Comment 35

[Akkana Peck]

We've had configurable key bindings for at least a year.  See
http://www.mozilla.org/unix/customizing.html for details.  No UI for it, but
it's better than modifying the jar files or pulling a source tree.

Comment 36

[Jesse Ruderman]

This regressed on Win32 when ben snuck in the patch for bug 108816.  Bug 
85526, "hitting backspace while using plug-ins goes back a page", might have 
also regressed.

Comment 37

[:deanis74]

Reopening due to bug 108816.

Comment 38

[Jonas Jørgensen]

Adding five keywords:

dataloss - it's easy to miss a form field while trying to focus it, and then
pressing backspace will go back, losing whatever you had in the form.

access - it's even easier for users with disabilities to miss the form field.

nsCatFood - users don't like losing what they typed in a form (I've heard
several IE users complain about that happening).

mozilla0.9.7 - this really should be fixed ASAP!

regression - that's what this is.

Changing priority to major since a topdataloss is _definately_ more than a minor
issue!

Comment 39

[Ben Goodger (use ben at mozilla dot org for email)]

This is not now, nor ever going to be fixed.

Reassigning to myself and closing as WONTFIX because having it assigned to
someone else will not make any difference whatsoever. 

Comment 40

[Ben Goodger (use ben at mozilla dot org for email)]

oops

Comment 41

[Ben Goodger (use ben at mozilla dot org for email)]

.

Comment 42

[Ben Goodger (use ben at mozilla dot org for email)]

.

Comment 43

[:deanis74]

Wow, I wish I could do that on bugs that I don't agree with.

Comment 44

[Jonas Jørgensen]

Gerv, you're the only person on the cc list with an @mozilla.org address. In 
other words, you're the only person here with more power than the Netscape 
people. Can't you do something?

And Ben, why did you remove the nsCatFood keyword? Many "normal" users hate 
this, since it will cause dataloss unless the user knows that you're supposed 
to press Forward to get your form data back.

Comment 45

[Gervase Markham [:gerv]]

I don't have "more power" than Netscape employees; just "different power". I'll
try and sort this out, but no promises.

Gerv

Comment 46

[Jonas Jørgensen]

Ben, I agree that it would be nice to have an easier shortcut than alt+arrow,
but backspace is simply a bad key for this. Can't we just use a key that doesn't
cause confusion and dataloss? Please?

Comment 47

[Aleksander Adamowski]

What about Ctrl+Backspace? No danger of dataloss here, still easy to press.

It's much more comfortable than Alt+LeftArrow, especially considering that with
some languages' keyboard layouts RightAlt is treated specially and generates a
different keycode.
Under Polish keyboard layout for example, RightAlt together with different
letters gives Polish diacritical characters, and to go back in Mozilla, one has
to press LeftAlt+LeftArrow, which either requires two hands or long fingers :-)

Comment 48

[Diego Biurrun]

*** Bug 120109 has been marked as a duplicate of this bug. ***

Comment 49

[Randell Jesup [:jesup] (needinfo me)]

Added dependency for bug 36922 which is for a competing version of what to do
for backspace - implement the ns <= 4.7 (and quite a few other browsers, and
gnus) behaviour: scroll up one page.  This has rather less dataloss possibility
than the IE compatible binding.  Also adding dependency on bug 57805 (front end
for user-defined mappings).

Comment 50

[John Levon]

*** Bug 125817 has been marked as a duplicate of this bug. ***

Comment 51

[John Levon]

invert sense of summary. Now :

backspace/delete is not mapped by default to go back

is WONTFIX, which makes sense.

Comment 52

[Jonas Jørgensen]

"backspace/delete is not mapped by default to go back" doesn't make sense since
backspace IS mapped by default to go back. This bug is about reversing it, and
it is WONTFIXed. Putting the original summary back.

Comment 53

[R.K.Aa.]

*** Bug 172956 has been marked as a duplicate of this bug. ***

Comment 54

[Robert Pollak]

I want to note that having the backspace key go back in history has become quite
annoying since we have "Type ahead find". It happens to me often that I want to
edit the search string and hit the key once too often.

Comment 55

[Owen Marshall (Not reading bugmail)]

Robert: See bug 177005. If you continue to have problems w/backspace & 
typeahead, comment there pls.

Comment 56

[Matt Beckett]

Just registered to report this.  Found it's a decade old bug that just couldn't seem to get traction enough for a fix (it's a default key binding, I assume it's not that it's hard to fix).
So it might only happen once a decade or so to an experienced user, but I have just lost a lot of webform data while trying to fix a typo... so yeah.  A decade later and it's still relevant.

Comment 57

[ziyunfei]

Chrome has disabled backspace going "back" in Chrome 52, so it's time to consider changing the default value of browser.backspace_action pref to 2 now.