bug#631: the M- notation suggestion

bug#631: the M- notation suggestion

[xah lee]

This is a suggestion on emacs's usability.

Recently in discussion at gnu.emacs.help, the subject of Alt+‹key›  
vs M-‹key› came up.

I think emacs's M-‹key› notation is one of emacs's a usability  
problem, contributing to its often cited big leaning curve.

Here's some detailed argument on why i think this should be chaged.  
Many emacs old users probably don't agree, but i thought it's good to  
send in one opinion anyhow.

---------------------

Emacs's M-‹key› Notation vs Alt+‹key› Notation

Xah Lee, 2008-07

Here're some reason i think emacs should adopt the Alt+‹key› and  
Ctrl+‹key› notation throughout its documentation. (as opposed to  
emacs's M-‹key› and C-‹key› notation)

UNIVERSALLY UNDERSTOOD

The Alt+‹key› or Ctrl+‹key› notation is universal among  
Windows and Linux. They account for about 95% of computers used word  
wide. Note that the word “Alt” and “Ctrl” are the exact labels  
printed on the Keys of PC Keyboards. PC Keyboards has probably more  
than 99% of market share.

IDENTICAL TO KEY'S LABEL

Using a notation that contains the actual label on keyboard's keys is  
much easier to understand. A beginning computer user, can read the  
“Ctrl+‹key›” notation and figure out which keys to press.  
Emacs's notation of “M-‹key›” and “C-‹key›” requires a  
learning step, even for experienced programers. Even though it is a  
minor one, but learning steps add up the complexity.

(Apple's computers, which account for about %4 marke share today,  
also use a notation where the name or symbol appears on the labels of  
Apple keyboard's keys exactly. (OSX's documentation uses the notaton  
“Command-‹key›” and “Option-‹key›”. Application's  
menus shows them as “⌘‹key›” and “⌥‹key›”. Both  
the word “Command” and symbol “⌘” appear on the key's label,  
same for “Option” and “⌥”.)
Meta Is Alt In Practice

By default on all major OSes in use (Windows and Linux and OSX),  
emacs maps its Meta to Alt key. So, practically speaking, the Meta  
key is the Alt key. (Aquamacs, perhaps the most widely used emacs  
distro on OSX, by default has Alt for Meta.)

KEYBOARDS DON'T HAVE META KEY TODAY

The Meta key was one of the modifier key on obsolete keyboards used  
by lisp machines in the 1980s. (for photos and detail, see: Why  
Emacs's Keyboard Shortcuts Are Painful)

There is practically no keyboard today that has the Meta key. Sun  
Microsystem's keyboard has a key labeled with a diamond “◆”.  
Sun's official documentation refers to this key as Meta key. (e.g.  
search http://docs.sun.com/ on “Meta key”.) Sun's keyboards have a  
market share perhaps less than 0.01%.

For photos and more commentary on Sun's keyboard, see Computer  
keyboards Gallery.

MISC FACTS

Historically, a “Meta+‹key›” shortcut in emacs can also be  
invoked by “Esc ‹key›” or “Ctrl+[ ‹key›”. The design  
was that way mostly because at the time, many terminals do not have  
or support the Meta key, and Terminal is a primary application in  
computer use in the 1980s. The other reason is that, in emacs's  
implementation, the Meta+‹key› is simply a ASCII control character  
sequence. Today, perhaps all terminal↗, console↗, Command line  
interface↗ apps support Meta as Alt either by default or in a  
preference setting.

The ability of pressing Esc for Meta might be still useful for some  
people. Users who needed that feature could easily read about it in  
emacs doc. (I myself used “Esc ‹key›” exclusively during  
1998-2004, mostly because it was a one-brainless solution that works  
on all telnet apps regardless of hardware, OS, or setup, and i  
frequently need to use different machine, OS, or remote servers.)

A argument from user interface perspective, is that multiple  
insignificant choices or options are not good because it increases  
complexity and causes the user to sidetrack their focus on tasks. KDE  
and Gnome, solved this problem for linuxes by adopting wholesale  
Microsoft Window's interface starting about 1998. (before KDE and  
Gnome, GUI apps on unix use a variety of “Windows Managers” that  
has incompatible User Interfaces, each claiming superiority.)

Note: Whether to use the “M-‹key›” or “Alt+‹key›”  
notation has little to do with “Esc ‹key›” feature.

PS Note that Microsoft Windows used to use the Alt-‹key› notation.  
Only in recent years they changed the minus sign to plus sign.  
Arguably, this is a good change because the plus sign better  
indicates key combination.

   Xah
∑ http://xahlee.org/

☄


☄

bug#631: the M- notation suggestion

[Richard M Stallman]

    The Alt+?key? or Ctrl+?key? notation is universal among  
    Windows and Linux.

I think you mean "GNU/Linux".  Linux has no user interface and its
documentation has no reason to refer to any specific keys.

Our convention for the GNU system is C-x and M-x, which is what you
are proposing to change.

    Using a notation that contains the actual label on keyboard's keys is  
    much easier to understand. A beginning computer user, can read the  
    ?Ctrl+?key?? notation and figure out which keys to press.  

There is some validity in that argument.

For the Emacs Manual I have a feeling this would be very wasteful.
But maybe it would be ok for some of the help commands, maybe even for
all of them.

bug#631: the M- notation suggestion

[xah lee]

> I think you mean "GNU/Linux".


Yes. GNU/Linux. Sorry i forgot i'm talking to you. LOL. (i do support  
GNU/Linux naming)

> Linux has no user interface and its
> documentation has no reason to refer to any specific keys.


In Gnome or KDE, when you pull menus, doesn't the shortcut shows  
beside the menu command names?

I think I forgot to mention this. I think if this suggestion is to be  
adapted, changing the shortcut notation as they are displayed in  
menus should be part of it. Perhaps even more important than the manual.


> Our convention for the GNU system is C-x and M-x, which is what you
> are proposing to change.
>
>     Using a notation that contains the actual label on keyboard's  
> keys is
>     much easier to understand. A beginning computer user, can read the
>     ?Ctrl+?key?? notation and figure out which keys to press.
>
> There is some validity in that argument.
>
> For the Emacs Manual I have a feeling this would be very wasteful.
> But maybe it would be ok for some of the help commands, maybe even for
> all of them.


Wasteful, as in too much effort and little benefit?

I think the effort would be few hours of interactive find/replace in  
the emacs manual's info source. O, there's also emacs tutorial info  
file. Possibly a few more.

the other part is to fix the elisp code so that menus display  
shortcuts using the Alt+ ant Ctrl+ notations. I'm not sure how this  
can be done or how much effort it takes.

I think the benefit would be great, because it gets rid of one big  
learning curve perception that stops people from using emacs.

   Xah
∑ http://xahlee.org/

☄

-------------------
On Jul 30, 2008, at 6:59 PM, Richard M Stallman wrote:

     The Alt+?key? or Ctrl+?key? notation is universal among
     Windows and Linux.

I think you mean "GNU/Linux".  Linux has no user interface and its
documentation has no reason to refer to any specific keys.

Our convention for the GNU system is C-x and M-x, which is what you
are proposing to change.

     Using a notation that contains the actual label on keyboard's  
keys is
     much easier to understand. A beginning computer user, can read the
     ?Ctrl+?key?? notation and figure out which keys to press.

There is some validity in that argument.

For the Emacs Manual I have a feeling this would be very wasteful.
But maybe it would be ok for some of the help commands, maybe even for
all of them.

☄

bug#631: the M- notation suggestion

[Lennart Borgman (gmail)]

Richard M Stallman wrote:
>     Using a notation that contains the actual label on keyboard's keys is  
>     much easier to understand. A beginning computer user, can read the  
>     ?Ctrl+?key?? notation and figure out which keys to press.  
> 
> There is some validity in that argument.
> 
> For the Emacs Manual I have a feeling this would be very wasteful.
> But maybe it would be ok for some of the help commands, maybe even for
> all of them.

Please put this on the TODO list. I think both the menus and the help 
command should reflect the actual keyboard labeling (for a standard 
keyboard on the used OS).

bug#631: the M- notation suggestion

[Richard M Stallman]

    > Linux has no user interface and its
    > documentation has no reason to refer to any specific keys.


    In Gnome or KDE, when you pull menus, doesn't the shortcut shows  
    beside the menu command names?

Yes, but neither GNOME nor KDE is part of Linux.
Linux is just the kernel.

bug#631: the M- notation suggestion

[Richard M Stallman]

    > For the Emacs Manual I have a feeling this would be very wasteful.
    > But maybe it would be ok for some of the help commands, maybe even for
    > all of them.


    Wasteful, as in too much effort and little benefit?

I mean reducing what's visible on a page, and wasting paper.
But someone could try it and see.

bug#631: the M- notation suggestion

[xah lee]

On Jul 31, 2008, at 3:01 PM, Richard M Stallman wrote:

For the Emacs Manual I have a feeling this would be very wasteful.
But maybe it would be ok for some of the help commands, maybe even for
all of them.


Xah wrote:
     Wasteful, as in too much effort and little benefit?

RMS wrote:
I mean reducing what's visible on a page, and wasting paper.
But someone could try it and see.


who's doing this?

and what about making the menu show the Alt+ and Ctrl+ notation? Is  
anyone in particular doing it?

PS give me a hint perhaps on how to do the latter. I might experiment  
on both.

   Xah
∑ http://xahlee.org/

☄

bug#631: the M- notation suggestion

[xah lee]

Richard M Stallman wrote:
«Linux has no user interface and its documentation has no reason to  
refer to any specific keys.  »

Xah Lee wrote:
«In Gnome or KDE, when you pull menus, doesn't the shortcut shows  
beside the menu command names?»

On Jul 31, 2008, at 3:01 PM, Richard M Stallman wrote:
«Yes, but neither GNOME nor KDE is part of Linux.  Linux is just the  
kernel.»


I was rather confused by the purpose or relevance of this remark for  
a total of estimated 60 seconds. (not a joke)

   Xah
∑ http://xahlee.org/

☄

bug#631: the M- notation suggestion

[Yavor Doganov]

Lennart Borgman (gmail) wrote:
> 
> I think both the menus and the help command should reflect the
> actual keyboard labeling (for a standard keyboard on the used OS).

IMHO this is close to impossible, since GNU/Linux runs on a variety
(at least 10) architectures, including archaic and modern machines
that have vastly different keyboards.  So there are many "standard
keyboards" for the OS GNU/Linux, also for the various free variants of
BSD.

"M-" existed since about forever; wiping it out will do more harm than
good.  If a new Emacs user has problems understanding it and finding
the right key on her keyboard, she surely will have much more problems
with other Emacs features, let alone more complicated concepts and
advanced usage.

Also, you should not consider only the Emacs manual.  Think of the
tens or hundreds of manuals of add-on packages, non-Emacs packages
(like Texinfo), and knowledge base like the mailing lists or sites
like the Emacs Wiki.  Changing something as fundamental as this for no
apparent benefit is a bad idea.  IMHO.

bug#631: the M- notation suggestion

[xah lee]

PS if someone send me url for the texinfo source for the emacs manual  
and emacs tutorial, i'll be happy to send back the modifier version,  
where any M- or C- notation is changed to Alt+ and Ctrl+, and any  
relevant place that might needs to be edited.

btw, the texinfo source for the emacs manual on the official gnu site  
may be outdated or bad. Two years ago i tried to use it and ends up  
spending several hours fruitless, finally got helped by given the url  
for a good copy of the texinfo source. (For detail see http:// 
xahlee.org/emacs/gnu_doc.html )

   Xah
∑ http://xahlee.org/

☄

bug#631: the M- notation suggestion

[xah lee]

Lennart Borgman (gmail) wrote:
«I think both the menus and the help command should reflect the  
actual keyboard labeling (for a standard keyboard on the used OS).»

On Jul 31, 2008, at 11:39 PM, Yavor Doganov wrote:

«
IMHO this is close to impossible, since GNU/Linux runs on a variety  
(at least 10) architectures, including archaic and modern machines  
that have vastly different keyboards.  So there are many "standard  
keyboards" for the OS GNU/Linux, also for the various free variants  
of BSD.
»

Microsoft Windows and GNU/Linux systems, together has about 95%  
market share of all computing systems. Their keyboard is typically PC  
keyboard, which has i estimate 99.9% market share world wide. Apple's  
computers use their own Keyboard but also has Alt and Ctrl printed on  
the keys.

So, practically speaking, this wouldn't effect those ~0.1% machines  
that emacs supports. For these users, typically they are advanced  
programers and they know what they are doing. (for example, if they  
want to browse the web, use a particular programing language, game,  
or software, they will typically find those not supported or under  
supported on their platform, but it is well expected.)

Yavor wrote:
«
"M-" existed since about forever; wiping it out will do more harm  
than good.  If a new Emacs user has problems understanding it and  
finding the right key on her keyboard, she surely will have much more  
problems with other Emacs features, let alone more complicated  
concepts and advanced usage.

Also, you should not consider only the Emacs manual.  Think of the  
tens or hundreds of manuals of add-on packages, non-Emacs packages  
(like Texinfo), and knowledge base like the mailing lists or sites  
like the Emacs Wiki.  Changing something as fundamental as this for  
no apparent benefit is a bad idea.  IMHO.
»

I think you are right that to do this completely would be near  
impossible, because that's over-riding some 25 or so years of emacs  
history. However, i think the benefit still outweight the negatives.

Also, changes that are few order of magnitude happens in commericial  
world often. A good example is Apple computer's switch from Motorola  
chip to PPC chip ~1995, and Mac OS to OSX ~2001, and PPC chip to  
Intel chip ~2006.

Some of these changes maintained some level of compatibility, but in  
general it wiped out several years of accumulated code,  
documentations, and world wide user expectations on how these  
software worked. Same happens in Microsoft's products. These  
commercial corps do that in order to survive.

Emacs does not have a survival problem, at least not in the sense of  
commercial software. Emacs also have small number developers, most on  
a voluntary basis. I htink the effort required in this change is  
relatively small, the benefit is arguably good because it reduces the  
number one complaint people have about emacs, namely being difficult  
to use or learn.

The issue, about the impact of this manual change, on past emacs  
related documents, esoteric systems, and old emac user's  
expectations, is neglectable i think, because software systems  
changes all the time with accumulated baggage. For a OpenSource  
example, Perl went from perl4 to a incompatible perl5 starting in  
1993, and it went on to thrive in the dot com era (~1998).

The proposed change doesn't actually effect elisp code. It is  
primarly esthetic in nature.

There's a large thread discussing this issue, at:

Newsgroups: gnu.emacs.help
Date: Thu, 24 Jul 2008 10:36:02 -0700 (PDT)
Subject: What does 'run' do in cperl-mode?

http://groups.google.com/group/gnu.emacs.help/browse_frm/thread/ 
5b81fcfd40d1f4ca/

After the first 5 or so messages, the rest 60 or so is about  
discussing M- vs Alt+ issue.

The debate is somewhat heated, but i think it hasn't degerated into  
bad usenet flame feast. Most disagree with the change. If would be  
great if those of you interested in the issue have a peek there.

   Xah
∑ http://xahlee.org/

☄

bug#631: the M- notation suggestion

[Lennart Borgman (gmail)]

Yavor Doganov wrote:
> Lennart Borgman (gmail) wrote:
>> I think both the menus and the help command should reflect the
>> actual keyboard labeling (for a standard keyboard on the used OS).
> 
> IMHO this is close to impossible, since GNU/Linux runs on a variety
> (at least 10) architectures, including archaic and modern machines
> that have vastly different keyboards.  So there are many "standard
> keyboards" for the OS GNU/Linux, also for the various free variants of
> BSD.

I am not suggesting that this should work for every keyboard, but it 
could work for the most common. In the other cases we could just stick 
to the current notation in help and menus.

> "M-" existed since about forever; wiping it out will do more harm than
> good.  If a new Emacs user has problems understanding it and finding
> the right key on her keyboard, she surely will have much more problems
> with other Emacs features, let alone more complicated concepts and
> advanced usage.

Why? I remember myself searching for the <copy> key for example.

> Also, you should not consider only the Emacs manual.  Think of the
> tens or hundreds of manuals of add-on packages, non-Emacs packages
> (like Texinfo), and knowledge base like the mailing lists or sites
> like the Emacs Wiki.  Changing something as fundamental as this for no
> apparent benefit is a bad idea.  IMHO.

I do not think it is realistic or useful to change the manuals. It is 
static. I also do not think the notation for adding key binding can be 
changed. However the dynamic bits in menus and help can perhaps be 
changed without a very big effort. That change can be useful for newcomers.

I think this could be handled by just changing some central code in 
Emacs, but I am not sure. Here is a list of what I think must be handled:

    Implement M- => Alt-

    - Change push_key_description
    - Increase #define KEY_DESCRIPTION_SIZE ((2 * 6) + 1 + 
(CHARACTERBITS / 3) + 1 + 1)
    - Add option for whether to do M- => Alt-
    - Add keyboard values
    - Add OS dependent default values for "Alt"
    - Change the macro that reads those key sequences and converts them 
to vector format

I might be missing a lot of things so please add to this list if you are 
interested.

bug#631: the M- notation suggestion

[Joe Wells]

xah lee <xah@xahlee.org> writes:

> The proposed change doesn't actually effect elisp code. It is primarly
> esthetic in nature.

This is not true.  There are many places in the Emacs Lisp code which
recognize the M- and C- notation.

First, there is the read syntax (I'm using Emacs 22.1):

  ?\M-A ⇒ 134217793
  ?\M-\C-b ⇒ 134217730

  "\M-A" ⇒ "\301" (yes, this is a bit different behavior for M-A)

Then, there is the convention of making symbol names with prefixes for
use in key bindings:

  M-f3
  M-mouse-1
  M-drag-mouse-2
  M-double-mouse-2

Then, there is the lovely kbd macro for use in key bindings:

  (kbd "C-M-<down>") ⇒ [C-M-down]

Then, there is the use of the M- and C- notation by edit-kbd-macro.

Then, there are the key-description, single-key-description, and
read-kbd-macro functions:

  (key-description [?\M-3 delete]) ⇒ "M-3 <delete>"

There is also the text-char-description function.

-- 
Joe


-- 
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.

bug#631: the M- notation suggestion

[xah lee]

The change suggested is only about in emacs manual, emacs tutorial,  
and in menus.

There is no proposal to change emacs lisp's keyboard macro or elisp  
functions.

   Xah
∑ http://xahlee.org/

☄

On Aug 4, 2008, at 4:11 AM, Joe Wells wrote:

xah lee <xah@xahlee.org> writes:

> The proposed change doesn't actually effect elisp code. It is primarly
> esthetic in nature.

This is not true.  There are many places in the Emacs Lisp code which
recognize the M- and C- notation.

First, there is the read syntax (I'm using Emacs 22.1):

   ?\M-A ⇒ 134217793
   ?\M-\C-b ⇒ 134217730

   "\M-A" ⇒ "\301" (yes, this is a bit different behavior for M-A)

Then, there is the convention of making symbol names with prefixes for
use in key bindings:

   M-f3
   M-mouse-1
   M-drag-mouse-2
   M-double-mouse-2

Then, there is the lovely kbd macro for use in key bindings:

   (kbd "C-M-<down>") ⇒ [C-M-down]

Then, there is the use of the M- and C- notation by edit-kbd-macro.

Then, there are the key-description, single-key-description, and
read-kbd-macro functions:

   (key-description [?\M-3 delete]) ⇒ "M-3 <delete>"

There is also the text-char-description function.

-- 
Joe


-- 
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


☄

bug#631: the M- notation suggestion

[Lennart Borgman (gmail)]

Joe Wells wrote:
> xah lee <xah@xahlee.org> writes:
> 
>> The proposed change doesn't actually effect elisp code. It is primarly
>> esthetic in nature.
> 
> This is not true.  There are many places in the Emacs Lisp code which
> recognize the M- and C- notation.

Thanks Joe for this list.

I am thinking about the possibility to just change what is shown in 
menus and help texts (see my earlier reply to Xah's bug report). Such a 
change would of course be optional, but might help new users.

With such a change there would of course be a steop for new users when 
they are just frustrated because they do not understand how to do key 
bindings, "why does not (kbd "Alt-<down>") work?", etc.

And the manual would probably better still use the current notation 
(with a note about the new notation) because it is static and is read on 
different platform. So everything would not be better, but the initial 
treshold would probably be lower.

Can you see any technical trouble with the changes I am suggesting could 
be done?


> First, there is the read syntax (I'm using Emacs 22.1):
> 
>   ?\M-A ⇒ 134217793
>   ?\M-\C-b ⇒ 134217730
> 
>   "\M-A" ⇒ "\301" (yes, this is a bit different behavior for M-A)
> 
> Then, there is the convention of making symbol names with prefixes for
> use in key bindings:
> 
>   M-f3
>   M-mouse-1
>   M-drag-mouse-2
>   M-double-mouse-2
> 
> Then, there is the lovely kbd macro for use in key bindings:
> 
>   (kbd "C-M-<down>") ⇒ [C-M-down]
> 
> Then, there is the use of the M- and C- notation by edit-kbd-macro.
> 
> Then, there are the key-description, single-key-description, and
> read-kbd-macro functions:
> 
>   (key-description [?\M-3 delete]) ⇒ "M-3 <delete>"
> 
> There is also the text-char-description function.
>