Re: [9front] dangerous Exit menu item in sub-rio

[igor@9lab.org]

Quoth Jacob Moody <moody@mail.posixcafe.org>:
> On 8/18/22 03:17, Steve Simon wrote:
> > maybe it still has, but mothra used to switch its cursor to a skull’n crossbones if Exit was selected - to indicate another B1 click was needed to confirm the exit, any other mouse button to cancel the exit.
> > 
> > -Steve
> 
> Yes mothra still has this behavior. I am not opposed to adding
> a confirmation step. When the first patch was discussed I took a survey
> and the leading idea seemed to just be to have it exit. I think some of
> this comes down to how you use subrios. For how I use them, a premature
> exit would not ruin my morning. But I can very much imagine someone having
> a workflow where that would be quite the damper on their morning.
> 
> It seems to me that people that get quite a bit of mileage out of subrios
> would prefer to have a confirmation exit step. For those that don't, one
> more mouse click is not a painful compromise.
[…]

Thanks for the option to `Exit` a sub-rio.  The comment about a
confirmation before exiting made me remeber the `Window Systems Should
Be Transparent` [1] paper by Rob Pike where he asserts that
``confirmations don't work'', here the relevant section:

``Mux doesn’t ask for confirmation to do risky things.  It always does
what the user asks.  The most obvious case is deleting a window.  The
protocol for deleting a window is a two-step process: select Delete,
then indicate which window.  The action can be aborted before picking
the window, so there is an element of safety built in; selecting
Delete does not immediately endanger any window.  This is one reason
why a window must be selected by a button click rather than just by
the position of the mouse.  Once in a while a window is accidentally
deleted, but this will happen whatever precautions are taken, and the
pain of recovering it must be weighed against the bother of dealing
with a complicated protocol for deletion.  (Presumably software has
some safeguards; editors will save their contents, for example.)
Confirmations are ignored in practice.  Once a user is at all familiar
with a system, the confirmations are executed without thinking about
them; the act of deletion becomes ‘click, click’ instead of ‘click,
read the message, think about it, click.''

Note how Pike argues that confirmations don't work, however,
dangerous operations like `Delete` are a two step process.

I have accidentally `Exit`ed from sub-rio and unfortunately it
does keep happening and I think one reason is because `Exit' 
is *not* a two-step process, another reason might be because
`Exit' might be buried in the menu when many windows are opened.

Consider a menu that looks something like this:

  New
 Move
Resize
Delete
 Hide
 Exit
Window1
Window2
Window3
...
WindowN

Because `Exit` is buried somewhere in the middle, trying to select
`Window1` or `Hide` sometimes results in accidentally selecting
`Exit`, exiting a sub-rio with many windows.

Why does `Delete` not apply to a sub-rio, i.e., why does selecting
`Delete` and then clicking on the gray background of sub-rio not close
the sub-rio?  That way it would remain a two-step process and there
is no need to add another menu option.

Another option is to move `Exit` to the very bottom, below all the
windows, e.g.:

  New
 Move
Resize
Delete
 Hide
Window1
Window2
Window3
...
WindowN
 Exit

That way one is less likely to click `Exit` when trying to `Hide` or
select a window.

While typing this email I believe moody mentioned to have implemented
a confirmation step similar to mothra. That is indeed another option
and one that most people are familiar with as well.

Cheers,
Igor