A tricky situation in malone bug 60995

[Martin Olsson]

Matthew Paul Thomas wrote:
 >
> A confirmation alert is usually the worst possible solution to any 
> design problem. People treat it as an interruption rather than as a 
> serious question. (Some horrid Web sites already do this, with 
> JavaScript alerts of the form "Are you sure you want to navigate away 
> from this page?")

I agree that confirmation is often a really bad solution. This is 
because of human "habituation". Who actually writes "rm thesis.txt" and 
then press "y" to actually do the delete? Everyone writes "rm -f 
thesis.txt" and then a split second later they go "nooooo".

This problem was fixed nicely on Windows using the "Recycle Bin" concept 
but on *nix there is no good standard solution yet afaik? Imagine what 
would happen if someone suddenly checked-in a change to the "rm" command 
that removed the "-f" option, and motivated this change by saying that 
"oh, but people are losing data so we must make it harder to loose 
data". And typically you have much more data in a file than you have in 
a form.

And to justify this crippled BACKSPACE key you still would have to 
explain why this is not a problem on Windows ("the main platform of 
ignorant computer users")? Why is it that Firefox on Windows still has 
this "really serious data loss problem"? Maybe it's because if someone 
made a change like this in Firefox on Windows people would be converting 
back IE "en masse". Of course, this is Linux so people don't have a 
choice (unless they want to go proprietary and use Opera).

Because, as long as Windows has BACKSPACE==BACK there will be tons of 
newbie users coming over that all expect BACKSPACE==BACK. Every single 
one of them will be *annoyed* and *confused* when they discover this. 
And if they actually understand that it's intentional, a large majority 
of these users will be *disappointed* as well.


>> ...
> There is no dataloss for Web sites that allow caching, but there is 
> dataloss for sites that use HTTPS, such as wiki.ubuntu.com.  

Yea, I agree that HTTPS/bank sites cannot cache their data.
I would prefer to have a dialog in this particular case, just like how 
there is existing dialogs for stuff like "do you want to resubmit your 
POST data?" and so on.


> Which leaves us with the other option: making accidentally going back 
> harder to do. Alt+Left instead of Backspace achieves this, but it seems 
> to be *too much* harder for some people.

The problem is not that Alt+Left is "harder", it's that A) people don't 
know about Alt+Left, and B) people are used to using BACKSPACE.

 >
 > One alternative would be to make "[" the shortcut key for Back. It
 > would still be possible to press it by accident when no text field was
 > focused, but much *less likely* than pressing Backspace by accident in
 > the same situation. And it would have another benefit that Backspace
 > does not have: an obvious counterpart key for Forward, "]".
 >


This would not fix anything, in fact; to some extent, this type of 
thinking is the very cause of our problem. Lots of people just randomly 
come up with new shortcut keys. Then people get used to having them and 
once these bindings are out there you cannot change them without losing 
users. The problem with this particular idea (of using "[") is that on 
any non-english keyboard the "[" key is really hard to use. On my 
keyboard I have hold AltGr and then press "8" to type a "[".

Anyway, my point is that key bindings are like public APIs. Even if they 
are quite bad you cannot just change them. Microsoft and Windows and 
long understood this but as Linux starts to get actual users it's 
important that Linux understands this too (for key bindings).

You cannot expect people to relearn stuff. Relearning takes effort and 
while it might not appear to be a big deal for someone who spends most 
of their day at a computer; for the casual (normal!) computer user, even 
small amounts of relearning makes the whole app unusable.




		Martin