On Sat, 09 Feb 2002 12:12:53 +0100 esteemed Jonas =?ISO-8859-1?Q?J=F8rgensen?= did 
hold forth thusly:
> > That would be a nice solution. 
> 
> ...which is exactly what 
> http://bugzilla.mozilla.org/show_bug.cgi?id=52821 is about.

See the comment I posted in that bug listing some ways to 
deal with this problem. I do not expect they will do it 
though. See how 39057 was closed with WONTFIX and its 
somewhat analogous. 
   http://bugzilla.mozilla.org/show_bug.cgi?id=39057

Look at Matthew Thomas' reasoning. I disagree with the reasoning, especially now 
that we have tabs. The act of closing Moz can take away dozens of loaded pages. 
It does for how I use Moz. I currently have 30 pages loaded in tabs. 
At the same time I _rarely_  shut down my browser. Ctrl-Q isn't even something
that I ever do. I use the upper right hand corner X button to shut down once
Moz has leaded enough memory to make it necessary.

Also, I'm closing literally dozens of tab pages in the course of a dayt and
and I don't want to use the mouse to do it since I don't want to move my hands
away from the keyboard. Having the browser close on me represents a considerable 
loss. Even if the browser was enhanced to work the way Opera does (Opera remembers 
all the windows that were open when you shut down and reopens them on start-up) that 
still doesn't let you restore back pages that may have expired out of existence or 
whose contents have changed (eg a web log where stuff scrolls off the bottom). 


So I disagree with his reasoning. See below for that reasoning from 39057:




------- Additional Comment #71 From Matthew Thomas (mpt) 2001-04-22 05:17 -------

I apologize for not getting to this bug sooner.

For the third and (hopefully) final time, this bug is invalid. Yes, I know it 
has a lot of dups. And I know that in Mozilla's current state, there are 
dataloss problems which an alert could prevent. But an alert is not the right 
way to fix those problems. Here's why.

Alerts are the user interface equivalent of telemarketing calls. It may sound 
incredible to those who haven't spent a lot of time watching people use 
computers, but human beings in general are chronically unwilling to pay 
attention to what alerts have to say. Often they will blindly cancel out of an 
alert without reading it at all -- especially on Windows, where many alerts are 
unfortunately given a close box so the user doesn't even have to read the text 
of the *buttons* (let alone the text of the alert itself). Because they 
interrupt the user's task, computer users try to banish alerts just as 
ferociously as experienced Web users try to get rid of popup windows.

The more alerts you produce, the less attention users pay to any of them (in 
your software or in anyone else's), and this causes problems when the computer 
really *does* need to inform the user of something important. For this reason, 
alerts should be used as seldom as possible. They should be a method of last 
resort.

Unfortunately, novice UI designers often regard alerts as a method of *first* 
resort for solving UI problems, especially since they are quite easy to 
implement.
A: `Our users are doing {x} a lot when they don't actually want to.' (In this
   case, quitting Mozilla.)
B: `Hmmmm ... Oh, I know, let's put up an alert asking if they're sure they
   want to do it.'
A: `Uh, but wouldn't that be annoying if you really *did* want to do it?'
B: `I suppose so ... I know, let's have a checkbox so people can turn off the
   alert if they don't need it!'
A: `Oh, ok then ... Let's do the alert thing, with a checkbox. And of course
   we'll need a checkbox in prefs to turn the alert back on again.'
MPT: `Oh no, not more prefs! ... <whine> <whine> <whine>'

You can see an example of this process in bug 49378.

Now, confirmation alerts in particular (which is what we're discussing here) 
should only be used when the user is at risk of considerable loss -- loss of 
data, loss of privacy, loss of security, loss of face, whatever. If you try to 
close an unsaved document in Composer, you're in danger of data loss -- so 
Composer asks you if you want to save the document. For the same reason, if you 
close an unsent message, Messenger asks if you want to save it as a draft. For 
each of those cases, a confirmation alert can and should be used -- but only 
because there is no better way of solving the problem.

But having a general alert, `Are you sure you want to exit Mozilla?', would be 
(and is, in 4.x for Windows, unless you've dulled yourself to the pain) 
incredibly annoying. To make things worse, it would work only for a short time 
-- until you started to instinctively hit the `Enter' key whenever the alert 
appeared. And this habit would likely spread to other alerts, in Mozilla and 
elsewhere, even alerts which you actually really should have cancelled out of. 
That would have the potential to cause just as much dataloss as the alert was 
intended to prevent in the first place.

Here too we have the `checkbox fallacy' -- the idea that a checkbox in the 
alert, to prevent it from appearing in the future, will solve the problem of 
some people not wanting the alert. But the problem we are trying to solve is 
not that only *some* people need the alert *all* of the time; rather, it is 
that *all* people need the alert only *some* of the time. Even the most 
experienced user will occasionally press Control+Q instead of Control+W by 
mistake. For that reason, a checkbox to turn the alert on/off *all* of the time 
is useless -- just as a similar checkbox was in the `Add Bookmark' dialog, for 
example (bug 68654).

So, if there are problems which result in loss from unexpected exiting of 
Mozilla, they should be filed as bugs/RFEs for ways of solving the individual 
problems, rather than plastering the general case with an alert and making the 
user experience worse every single time they exit.
*   If accidentally exiting Mozilla, then starting it up again, takes too long,
    then the dependencies of bug 7251 should be fixed.
*   As Jesse said on 2000-09-28, people often close browser windows
    accidentally while entering a form (e.g. a message in a Webmail service).
    But having an alert on Mozilla exit wouldn't fix that problem, since it's
    a problem with closing any given window (rather than with closing the whole
    app). For that, bug 48333 should be fixed.
*   Since exiting the app from one window closes seemingly unrelated windows,
    bug 65121 should be fixed (as it has been in Internet Explorer for quite
    some time).

All of these particular problems can be solved by methods which are more 
pleasant (and as a result, more reliable) than asking the user every time if 
they're sure they want to exit. Presenting the user with an alert every time 
would be overkill; it would imply arrogance on the part of the Mozilla 
application; it would imply stupidity on the part of the user; and it would 
make every other alert which the user sees (in Mozilla and elsewhere) less 
effective. For those reasons, I will not allow it.

 

Reply via email to