Directory drill-down navigation

[joturner <joturner vt edu>]

After using Gnome 2.5.5 (which I believe uses Nautilus 2.5.7) for a while I've 
begun to grow accustomed to the spatial navigation.  While I like it, I also 
think it's worth mentioning spacial single-window navigation.

The effect now seems very similar to how the BeOS Tracker worked.  Every new 
directory was a new window.  For years people used this, and those who didn't 
appreciate it tried to make alternatives.  Eventually, the effort to overhaul 
the tracker became an open source movement sponsored by Be, Inc. called 
OpenTracker.  No doubt this has already been covered.  I mention it only to 
say that they moved from a true spatial navigation to a single-windowed type.  
Microsoft recently did something similar with some of the choices they took in 
XP.

I can see how such a change might cause a few to believe it's a slippery 
slope.  If you add the option for single-window drill-down, you have to put 
forward and back buttons, or else you've lost your ability to return to 
previous directories.  Once you add forward and back buttons, why not up?  And 
once up, why not a tree view to easily navigate between directories?  Now it's 
not spatial at all.

In defense of the suggestion I can only say that strict spatial navigation can 
have trouble with Fitt's Law.  (ref: 
http://www.cs.umd.edu/class/fall2002/cmsc838s/tichi/fitts.html)

Let's say you open the home folder by clicking on "[your user name]'s Home".  
The window opens away from the cursor, so you move the mouse over to it and 
double-click another directory.  The new window opens away from the old one, 
so you move your mouse again.  In my sample now, on my machine, the desktop 
icons are obscured, as is about 60% of the desktop, and I've travelled the 
mouse almost round-trip back to where I originally clicked the first desktop 
icon.  To see the desktop icons I'll have to move the highest folder window 
out of the way.

The beauty of spatial navigation is its simplicity.  I think that's pretty 
obvious.  It would follow, then, that a solution to the Fitt's distances would 
maintain that simplicity.

A few solutions have already been mentioned:

* Keep the same window.  The OpenTracker approach was to put back/forward/up 
buttons on the window.

* Have an option to close the old window when a new one is opened.  This 
option seems to have the same Fitt's distances as the current spatial view 
without the ability to navigate to previous directories.  It does, however, 
allow the user to automatically close windows so that windows or icons below 
are not obscured.

* Keep spatial navigation the same, but offer a way to change the default 
navigation to explorer-like navigation.  This doesn't fix spatial navigation, 
but it does offer a way for users who do not wish to use to avoid it.

These may have been mentioned before I joined, but I'll list them here 
(apologies, I wish I knew who to give credit for for these)

* Use tabs, and create new ones during drill-down.  This would work well until 
the number of tabs became too many and Fitt's Law began to break down.  Too 
many tabs, also, would be cumbersome and easy to lose track of.

* Create new tabs by request of the user.  This may suit a more advanced user, 
but a user unaccustomed to this type of customized navigation would not be 
able to navigate backwards through previous folders.

* Instead of tabs, create a back-stepping list on the side, a "history" of 
previously visited directories.  This would remove the back/forward/up 
buttons, but it adds another element that the user would have to understand.

* Add keyboard accelerators to jump back and forth between open spacial 
folders.  This is an advanced user activity, but it does solve the Fitt's Law 
distance, assuming, that is, that there aren't more than a handful of open 
spacial folder windows open.

I'm sure there are other solutions as well, each with their own set of 
benefits and drawbacks.

One thing is for sure, though, spacial navigation in Nautilus in its current 
form has been attempted by Microsoft and BeOS, and both have since moved to a 
single-window drill-down.  For some users the single-window approach is more 
comfortable, and from a Fitt's Law standpoint, it is more compliant.  This 
doesn't mean the spatial navigation should be made more complicated to 
compensate, but a simple solution should be found and applied, something that 
is easy for the new user to learn while maintaining usability for the more 
advanced user (as well as the user that uses the computer for long periods of 
time where Fitt's Law plays a big part).

Jonathan