Re: A Violent Realisation [Was: Preferences]

[Havoc Pennington <hp redhat com>]

Dick Porter <dick ximian com> writes: 
> Most of the complaints seem to be about _removing_ options that exist
> already in GNOME 1.4, with the explanation being that it simplifies things
> for new users this way.

The problem is that it's true - it does simplify things for new
users. And it also avoids the other costs of excess prefs I listed. 

Here are the ones I listed:

 - Too many preferences means you can't find any of them.
 - Preferences really substantively damage QA and testing.
 - Preferences make integration and good UI difficult. 
 - Preferences keep people from fixing real bugs.
 - Preferences can confuse many users. 
 - The preferences dialog is finite in size. 

Taking viewports specifically, here are the costs of having both
viewports and desktops:

 - having both viewports and desktops makes the 
   prefs dialog for the pager much larger than it currently 
   is, and requires people to understand a fairly hard-to-explain 
   concept.
 - having both viewports and desktops makes a lot of subtle UI
   behaviors (like what happens when you reopen a dialog that's
   already open on another viewport/desktop, or how Xinerama/multihead
   work) harder to get right
 - having both viewports and desktops adds substantial implementation 
   complexity to both the pager and the window manager
 - having both viewports and desktops traditionally means that there's
   an odd and random division of features between the two (viewports
   have certain keybindings or edge flipping, while desktops don't, 
   purely for implementation laziness reasons)
 - having the choice confuses many users

Now the benefits of having both viewports and desktops:

 - you can have a double-nested grid with a grid of viewports
   inside a grid of desktops.
 - users can choose whether they want windows to overlap 
   between work areas.

So that's a concrete analysis. My view of the concrete analysis is
"cost outweighs benefit" - but sure, I can change my mind, given
logical additions to the concrete analysis, instead of arguments that
all preferences that have ever existed should still exist.
 
Note that "windows overlap between work areas" is a benefit of
_viewports_, not a benefit of _having both desktops and viewports_. 
Switching to viewports as the one thing we have is another option, 
instead of adding a preference.

> I dont think you can legislate for what is or is not a suitable feature or
> preference: you have to evaluate each suggestion or patch on its merits.
> (Merits including of course how badly it impacts the code, as well as
> usability or aesthetic grounds.)

Sure, I would love to see arguments on each preference. The thing is
that people tend to argue "well preferences don't hurt anything,"
which I believe is patently untrue. They have a cost, they have a
benefit; you have to weigh it and give a _good_ justification for each
preference.

People love to complain about missing prefs, but don't want to do the
hard work of deciding which to have. ;-)

Take gnome-terminal for example, I feel it has too many settings
already, and I'm out of prefs dialog space, but if I added all the
settings that have been requested in bugzilla, it would have twice as
many as it already does.

And none of the people asking for more prefs are willing to make the
hard global decisions about which existing pref should die to make
room for their new pref.

So maintainers have to have the guts to a) understand user complaints
as indicating a real problem but b) not just blindly add prefs as a
way to address those complaints. There are often other solutions.

Havoc