Re: Nautilus 2.6 - We're going all spatial

[Seth Nickell <seth gnome org>]

On Mon, 2003-09-15 at 13:10, Ettore Perazzoli wrote:
> On Mon, 2003-09-15 at 10:28, Dave Camp wrote:
> > For the 2.6 cycle, the nautilus crew is trying out a new UI that
> > should give us the best of both worlds.  The idea is present an object
> > oriented UI from the desktop, but to allow users to open navigation
> > windows if they prefer them.  This means that opening a folder from
> > the desktop will give you an object window.  Opening folders from
> > object windows will give you new object windows.  You can right click
> > on a folder from an object window and select "Navigate Folder"[1], and
> > get a normal nautilus window.
> 
> Is this really a step forward?
> 
>       * It is not at all clear to me that the spatial model is more
>         usable than the browser model.  (Sure, it makes a lot of sense
>         and is more real-world-like, but as it's being pointed out
>         before it does have drawbacks.)

Abstraction (or Cooper's related concept of "cognitive friction") causes
people more trouble using computers, overall, than a lack of real-world
mapping. There's definite benefit to enabling positive transfer (the
primary benefit of real-world mapping), but primarily in the learning
phase. Being able to build a working model of how something functions
will permanently effect how a user works with an artifact. 

Natural aptitude at coping with abstraction (sometimes even a desire to
deal with abstraction) is, I believe, one of the primary qualities that
allows somebody to be a competent programmer. OTOH, most people have
difficulty both learning and working inside of layers of abstraction. I
would say this is the principal cognitive difference, on average,
between developers and users.

Using a navigation ("browser") model means:

1) People will not understand Nautilus as well and will instead rely on
memorized scripts for accomplishing tasks

Its less likely that somebody will develop a conceptual model that
matches the design model when its mediated through a system image with
more indirection. Many people will never develop one and will instead
memorize scripts to accomplish their common tasks, which will cause them
trouble in the long run. A script is a useful cognitive tool, but breaks
down much more quickly when trying to do a new task, the environment
changes, or something breaks.

Spatial model (more typically called object model unless you are an Ars
Technica writer ;-) says "here is an object". Navigation model is more
precisely called a mediated navigation model (the "mediated" part being
the crux of the problem) and forces you to add "here is a viewer, it
lets you visualize an underlying set of objects in different ways". This
adds a layer of indirection to a commonly used piece of the desktop.

2) Many of the people who do understand Nautilus will use it less for
simple tasks

Even if they develop a conceptual model that mirrors the design model
closely, a more sophisticated model will slow people down when
performing even relatively common tasks. Worse, it will decrease the
likelihood of using the tool on an incident by incident basis. Most
people won't be aware of this consciously, but if you measure people
operating under more complicated conceptual models of a tool you find
they are less likely to use it (even if at a higher level they can tell
you how to use it for all sorts of things, they just won't because it
imposes too much of a cognitive burden).

>       * You are actually making the Nautilus model more complicated (not
>         simpler) by exposing the user to two completely different kinds
>         of windows, for "object" and "navigation" purposes.  This seems
>         to defeat the basic premise of making the model easier to learn.

It makes the Nautilus model more complicated for users who choose to
work through the navigator.

>       * If "navigation mode" is only available from the menu bar or a
>         right click menu and everything on the desktop opens in "object"
>         mode by default, then users who prefer the navigation mode
>         (which might even be the majority of them :-)) 

I don't believe its going to be anywhere near a majority (I would expect
it to be substantially less than 10%) if you're looking at using GNOME
as any sort of corporate desktop. Furthermore, I believe there's very
strong overlap between people who prefer navigation mode and people who
don't and won't use Nautilus to a significant extent anyway, preferring
the terminal.

> are going to
>         suffer a lot, since they won't be able to just double-click
>         desktop icons anymore.

Once you're browsing inside a navigator window its going to stay that
way. Clicking on folders inside it is going to change the current
location not popup the folder.

> > This interface is partially inspired by the interface described in
> > http://arstechnica.com/paedia/f/finder/finder-1.html .  Interested
> > parties should read that before getting involved in the discussion.
> 
> I find most of the arguments in that article to be very subjective, and
> not very well-founded.

I agree. I put very little weight in what the writers at Ars Technica
think. I suspect Dave linked to it because it helps explains the MacOS 9
Finder's design and not because its a good defense of those ideas.

> Before making such a big paradigm shift it would probably be better to
> do a thorough research on how the model actually affects users. 
> (Research which you might have done already -- in which case it would be
> nice to know the results of that.  ;-))

Dave and I have been talking about this on and off for quite some time
now. The general principles operating behind this shift in model are
pretty much "design 101", not bleeding edge HCI hypotheses. There *are*
major trade-offs, particularly for technically proficient users
(everyone on this list). Less abstract designs often have this tendency.
Still, many of the people on this list are human and may find benefit
from this design even if they aren't aware of it (this is a surprisingly
common phenomenon). For non technically proficient users (many of whom
don't even really understand recursively nested folders, filesystem
paths, etc) this model is a no-brainer.

What Dave has tried to do is to switch the primary nautilus interface to
an object model while still providing for users who really want
navigation. I think there are some confusing aspects to the way this has
been done in the existing branch, but overall I think this is a pretty
sound approach.

> > Our general opinion coming out of that thread was that the Object
> > Oriented metaphor was probably easier to learn, and built a stronger
> > conceptual model for users, but that the convenience benefits of a
> > navigation window outweighed those.
> 
> So why are you making the object oriented metaphor the default?  I doubt
> there is really a learnability problem at all -- if the browser model
> were so difficult to understand people wouldn't even be able to browse
> the web (or use a file selector dialog).

Learnability in this case plays serious second fiddle to providing a
system image that helps people develop a useful conceptual model.
Learning is once, conceptual model will influence how (and how much, and
for what, and when, etc) the user works with Nautilus and the GNOME
desktop as a whole by virtue of Nautilus' pivotal role.

The browser model is substantially harder to use, and the browsing model
of the web is one of the things that makes it most tricky for building a
useful conceptual model. This was probably the largest design trade-off
that the web made (for technical reasons, and because the web is a
fundamentally different problem from navigating a smaller set of objects
that are already in a tree structure). The design flaws (even if
necessitated as part of an overall package of tradeoffs they are still
flaws) of the web should not serve a pattern to be replicated :-P

-Seth