[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kfm-devel
Subject:    Re: TreeMaps for Konqi?
From:       Keunwoo Lee <klee () cs ! washington ! edu>
Date:       2001-02-26 23:01:14
[Download RAW message or body]

On Mon, 26 Feb 2001, Marc Mutz wrote:

> This view is just cool. It has it's special uses, because it allows you
> to view thousands of items simultaneously. That's it's main feature. But
> most of all, it's just cool. It's like having a nice looking splash
> screen: It _may_ be good for something, but in general it's there to
> look good.
> OTOH, this view is new and maybe powerful. It's _very_ close to current
> research. Why do not have something that the Windows Explorer does not
> have yet?

My $0.02: judging from the screenshots, SequoiaView has severe user
interface problems.  I agree that it's a cool idea, but the visualization
is confusing and of dubious value.  The fact that it's relatively new
research from the lab (actually, the first paper was in '91, so it's not
that new) has its good points (novelty value) and its bad points
(usability unclear).

"Windows doesn't do it" also does not strike me as a particularly good
reason to add this to Konqueror.

OTOH, if somebody can smooth out the rough edges and figure out a way to
make the interface more intuitive, it might be OK.  Some friendly advice
for improving the treemaps, if you decide to implement them:

  + Don't use flashy metallic gradients.  They're distracting.  The
simpler color-coded versions at the following URL:

    http://www.cs.umd.edu/hcil/treemaps/

are much easier to understand.

  + Figure out a sensible way to organize horizontal vs. vertical
subdivisions.  The ad hoc subdivisions along both axes produce some very
difficult to understand visualizations, and (as the inventor notes) breaks
a basic rule of information design: Don't represent linear quantities with
areas.

  + Read all the technical reports from UMD's HCIL, linked from the above
URL, and related papers.  They seem to have done further research on the
usability of treemaps, and variants on the original layout algorithm.

BTW, if you really want to create a useful tool for figuring out how much
space a file takes up, I suggest you write an application (or KPart or
whatever) that queries the filesystem and presents a standard bar-graph
visualization, broken down by user-selectable criteria (directory, file
extension, regexp).  This would probably be easier to understand, easier
to code, and more generalizable.  Though it would not be as flashy, and
maybe not as exciting to program as treemaps :).

Finally, if you like cool visualizations, the URL above references a
contest at CHI97:

http://www.acm.org/sigchi/chi97/proceedings/panel/kem.htm

where a bunch of visualization and browsing technologies were pitted
against each other.  You might find one that you think is cooler than
Treemaps, and maybe more usable.

~k.lee

[prev in list] [next in list] [prev in thread] [next in thread]