Re: Nautilus 2.6 - We're going all spatial

[Seth Nickell <seth gnome org>]

> Also, I am not sure which MacOS users you mean here.  MacOS 9 used the
> spatial model, MacOS X doesn't.  So which MacOS users (9 or X) have a
> better grasp on the model than Windows users?
> 
> Do fewer MacOS X users understand the file system than MacOS 9 users
> did?  That would be interesting to know; in the context of discussion,
> the OS X vs. OS 9 comparisons seems much more meaningful than the
> Windows vs. OS 9 comparison.

First, AFAIK based on little tidbits of information, it sounds like the
OS/X finder might actually be moving more toward an object model. But
that aside... That would be an interesting comparison.

> > People are *capable* of handling some degree abstraction, and it is
> > necessary in many places. I think perhaps you are imposing "can use",
> > "can't use" as binary conditions that are too strong. 
> 
> I am just saying -- the browser model is inherently part of the
> experience of using a computer.  You are describing it as evil; it
> isn't.

The browser model is simply more abstract. That gives it inherent
(major) downsides. Think of it as adding more bureaucratic red-tape: it
doesn't make things impossible it just means people are going to do it
less. These downsides can be overridden by other concerns (using our
red-tape metaphor, bureaucracy can be important for protecting dangerous
procedures, diffusing power, etc, but you wouldn't want it getting in
the way of walking the dog). It is easier to outweigh the downsides when
the operation you are making abstract can be ignored in favour of a
reliable cognitive script.

> > > > 2) Many of the people who do understand Nautilus will use it less for
> > > > simple tasks
> > > 
> > > Have you tested this?
> > 
> > User testing is not a particularly good technique for determining this.
> > It is simply one HCI technique out of many (when its useful its a good
> > technique because its comparatively easy, cheap, and fast). A long-term
> > use study would be much more appropriate. Have I conducted a long-term
> > use study? No.
> 
> Then where is (2) coming from?  Is it just a guess?

It is not a guess, at least not any more 99% of the working choices made
by programmers are guesses.  People with domain expertise, even faced
with new problems in their domain (every problem is a new problem), are
not making guesses (at least not with the associated negative
connotations) even when they do not conduct "scientific" tests.

It is coming from a trained-designer-with-a-formal-background-in-HCI's
analysis of the problem. This analysis is formed based on my knowledge
of the general effects of increase in cognitive friction on people's
behavior patterns in conjunction with a general sense of the problem
domain.

> IIRC (and my memory might be failing me, I guess) Windows 95's initial
> model was mostly spatial.  If you double clicked on an object, it would
> open a window for it.  If you double clicked on the same object again,
> it would give you the same window.

> Of course it had flaws and there were other ways you could things and
> have multiple windows for the same object, but that was still the basic
> model and how people used it most of the time.  The non-spatialness-ness
> aspects of it weren't even visible to the naive user.  (And we are
> focusing on the naive user, right?)

1) It was too visibly *not* the Object model to allow people to form a
solid conceptual model. Conceptual model's are fragile things. If people
find evidence that a model doesn't hold don't hold they will often throw
it away (or worse, will hold them to the exclusion of developing a new
model, but not have any confidence in them).

2) Windows provided certain artifacts that you would also find in an
object model but they did not manage to communicate the design model
with sufficient force to develop a parallel conceptual model.

Models are not so much about features as they are about communication by
way of the system image. They are the developers communicating a certain
way of understanding the function and operation of the software to the
user. Windows gave both a mixed message AND provided contradictory
information. Communication didn't succeed in many cases.

> > As I said, design is a series of tradeoffs. Even if a
> > certain choice was *justified*, this doesn't mean that its not a flaw.
> 
> Of course a design is a series of tradeoffs -- in fact, when you pick
> the spatial model you are living with the flaw that a common operation
> (picking a file to open) has a side effect (many windows open) that we
> *know* many, many users reported as annoying.

But what percentage of users? There are many upsides to that "side
effect" too, especially for people working out of document folders (the
norm). It is mostly a downside for people who do lots of tree-walking
(programmers, sysadmins). This is not common behavior with most office
users. I expect the most likely place for people to go in the file tree
after opening a folder is back to the parent folder, so this behavior
makes the "common case" very easy. It really breaks down for people who
deal with deep and complex trees, I agree, which is why we provide
alternate methods for these special use cases (navigation browsing, and
perhaps a modifier key to not leave the parent folder open).

-Seth