Re: Experimentally unbind M-o on the trunk

[Eli Zaretskii]

> From: Yuri Khan <yuri.v.khan@gmail.com>
> Date: Fri, 12 Feb 2021 00:53:46 +0700
> Cc: Matt Armstrong <matt@rfc20.org>, Lars Magne Ingebrigtsen 
> <larsi@gnus.org>, "Alfred M. Szmidt" <ams@gnu.org>, 
>       Gregory Heytings <gregory@heytings.org>, Jean Louis <bugs@gnu.support>, 
>       Emacs developers <emacs-devel@gnu.org>
> 
> > ??? Typing '?' or C-h in the middle of any key sequence should (and
> > usually does) provide discovery.
> 
> Not really. In a fresh ‘emacs -Q’, when I press ‘C-x ?’, I get a help
> buffer that has three logical pages and is 10.5 screenfuls long.

Yes, and that's bad because...?

Of course, any patches that would succeed in somehow guessing what the
user is after, and presenting the alternatives in a smarter order --
any such patches will be very welcome.  But saying that there's no
discovery at all is a bit extreme, don't you think?

> In the discovery scenario, which supposes I don’t know what I’m
> looking for but will know when I see it, I have to scroll through
> all that until I find the ‘C-x v d’ binding I was looking for. It
> says, literally:
> 
>     C-x v d    vc-dir
> 
> so if I don’t already know vc- means version control, I may not
> recognize that command as the one I want.

It's easy to find examples that prove your point if you really want
to.  But similar issues happen with menus, although menu items are
usually longer, so such issues are more rare.  But they still exist.
For example, what is "Tools->Read Net News" about? are you sure a user
who doesn't know about Gnus will immediately understand what that
does?  Or how ab out "Edit->Paste from Kill menu"? or
"Options->Customize Emacs->Faces Matching...".  Etc. etc. -- lack of
basic knowledge always requires some kind of bootstrap-like process
until you know enough of the basics to be able to tell in advance what
you need to look for.

> Also, importantly, when I ask for help, I cannot proceed to
> immediately execute the command I found, nor continue the discovery. I
> have to re-type the sequence from the start.

You lost me here.  The list popped up when you type '?' shows what you
should type -- that's "proceed to immediately execute the command you
found".  You should retype it, yes, but what would you suggest
instead? ask the user to remember what was typed before the question
mark?

> (I can also click on the blue underlined text, which does not
> execute the command but displays extended help on it. Thanks, Emacs,
> how do I go back?

Back from where?  You didn't go anywhere, you were just shown help.
Just continue what you were typing.  Emacs isn't modal, at least not
normally, so "going back" from what it pops up makes little sense.

> There’s no ← icon on the toolbar

But there's that "[back]" button at the end of the help text...

> and the History Back button on my mouse causes a “<mouse-8> is
> undefined” message in the echo area, and the context menu mouse
> button marks a region. (Out-of-character: Yes, I know about ‘l’ and
> clicking the (Help) button in the modeline. My character doesn’t.))

Are we still discussing discoverability, or are we venting?  How is
the fact that your system has a misconfigured mouse relevant to the
issue at hand -- which was discoverability?

And again: patches to improve the UX are more than welcome.

> Compare with the menu: Tools (which, while looking suspiciously like
> the “everything else” bin, is likely to contain the functionality I
> want) → Version Control (yes!) → VC Dir (somewhat poorly named but
> okay). *This* is discovery.

Of course! you've cooked the example, so of course it serves you well.

Now try "C-x d" and tell me where you find some operation.

> On a semi-related note, how do I go back up from a submenu in ‘tmm-menubar’?

Why should anyone use tmm-menubar when we have menus on all types of
displays?  (And no, I don't know, I never used tmm-menubar.)