Re: Trying a new toolbar style

[Shaun McCance <shaunm gnome org>]

On Wed, 2009-07-29 at 10:27 -0400, Dan Winship wrote:
> On 07/29/2009 09:51 AM, Johannes Schmid wrote:
> > Hmm, yes that definitely looks different than I expected. I am not
> > entirely sure though if it makes sense to have labels in the toolbar at
> > all because it would be way better if icons would be self-explaining.
> 
> Right. The idea isn't "label the confusing icons and don't label the
> good ones". You're supposed to label the important/frequently-used
> icons, thereby (a) making them easier-to-notice by being stylistically
> different from the less important icons, (b) making them
> easier-to-identify by having two different ways to recognize them
> instead of only one, and (c) making them easier-to-click by making them
> bigger, per Fitts's Law.

Does it really accomplish (c)?  The classical formulation
of Fitts's Law only deals with motion in one dimension.
In 2D interfaces, it effectively ignores time introduced
due to variance in the direction of movement.

http://en.wikipedia.org/wiki/Fitts%27s_law

For most starting locations on the screen, having text
beside the icon adds very little to the "width" of the
object to be acquired.  Text below icons adds much more
to the "width".

What text beside icons does is give you a greater tolerance
in the direction you send your mouse.  Note, however, that
angular tolerance decreases with distance from the object.

Anyway, I'll stop short of writing mathematical formulas
that make everybody's eyes gloss over.  I'm just saying
that if the reason for doing something is to make a target
easier to acquire, we should have some reasonably assurance
that it actually accomplishes that goal.

--
Shaun