Improvement to the consistency of scale effect

[fermat]

Binary package hint: compiz

Description: Ubuntu 10.04 LTS
Release: 10.04
compiz:
  Installed: 1:0.8.4-0ubuntu15
  Candidate: 1:0.8.4-0ubuntu15
  Version table:
 *** 1:0.8.4-0ubuntu15 0
        500 http://gb.archive.ubuntu.com/ubuntu/ lucid/main Packages
        100 /var/lib/dpkg/status

This is an enhancement/usability improvement request to make scale effect more consistent. I have compared the scale effects of mac os and compiz (no flames please!) as I use both during the day.I always felt something was amiss when I used the compiz scale whilst I felt much more comfortable with the mac version. Both scale versions improve the productivity dramatically.
I did a small HCI 'study' with test subjects being myself and 6 colleagues for three weeks. I discovered the following:

1.We are better off and more efficient in remembering the relative positions
(scaled vs unscaled) if they are either in the same scaled location or in the same quadrant irrespective of intermediate changes to other windows. If their position is changed, the scale obviously needs to place it in the appropriate location.

2.The 'grid' scaling (used in compiz and in mac os snow leopard) is inefficient as it forces the muscle memory to relearn the positions or to search for them in the 'grid'. Instead, a scale effect maintaining the relative spaces and positions and sizes gives the brain a consistent map. Maintaining the relative sizes of windows enables easy identification. The move from this, to the grid/same size method in mac os caused so much gripe and many (including me) created hacks to use the old effect.

3. Linear motions are better than curved ones; while the compiz scale effect is elastic and window movements are 'curved' creating a pleasing effect, it puts more strain to track a window that is partly visible in normal mode to its scaled mode. 'Straight' movement from normal to scale seems to improve the tracking abilities.

I have attached a series of screenshots to show how mac and compiz scales windows.

Mac
1.Start - normal set of windows with varying overlaps/focus and size
2.First scale - observe the positioning of the scaled windows
3.Change Focus - Now i change the focus to another window
4.second scale - the positions are the same (not in the case of compiz)
5.Add window - I added another window
6.third scale - the relative positions are ore or less the same for windows away from the new window, the windows near the new window tend to stay in the same quadrant

For all the scale effects, the background darkened quickly , reducing clutter.

Compiz
1.First - normal set of windows (don't know why the screengrabber shows itself!! an enhancement request for that may be?)
2.First scale - normal scaling, note the position of windows and the grid behaviour
3.Close window - i had closed the windows on the bottom right
4.scale after close window - note how the position of the nautilus with the saved screenshots has moved from third to fourth quadrant)
5.add window - added emacs window
6.scale after add window - note again, the change of window positions (especially the screenshots nautilus window)

For all the scale effects, the background darkened (if enabled) as a fade-out, looking pretty but distracting.

I propose the following changes to compiz:
1.option to modify the 'elasticity' and 'bounciness' of window movement during scale
2.consistent placement of windows (through observation 1,2)
3.ability to darken the background quickly (or provide an option to change the speed)

The screenshots may not provide all the necessary data, and I have included only few for brevity. I can provide more data if needed.

[quequotion (quequotion)]

I hate to see such effort dismissed with a simple "invalid" (however technically correct as this is a feature request, not a bug report). Post this in the compiz forums if you haven't already. Compiz can always use a few functionality-minded testers/developers. There are a lot of great features, including productivity enhacements, but I too feel some of them could benefit from serious human-computer interaction research.