Lets take the worst of windows and make it the unchangeable default of linux

Nils Philippsen nphilipp at redhat.com
Thu Feb 19 15:29:37 UTC 2004


Bah, resending due to mailman being an ass to PGP/MIME signed email.

---------------------------------------------------------------------

On Wed, 2004-02-18 at 17:40, Alexander Larsson wrote:
> On Wed, 2004-02-18 at 16:25, Alan Cox wrote:
> > On Mer, Chw 18, 2004 at 04:05:56PM +0100, Alexander Larsson wrote:
> > > This has been discussed a billion times before. I'm not having
this
> > > argument again. If you want to see the rationale for it, read for
> > > instance http://www.ometer.com/free-software-ui.html or google for
some
> > > of the discussions about this. 
> > 
> > Or go read some stuff on both UI design _AND_ on product selection
attributes.
> > The latter of which is an established theory driving billions of
dollars of
> > marketing research and which in part strongly disagrees with the
claims of
> > that article.
> >
> > More specifically there are sets of attributes people use to pick
product or
> > service. Some of those attributes are more important than others.
Certain
> > attributes are sufficiently important that the user will avoid the
product
> > or switch given the opportunity. Others matter a lot to a user but
aren't
> > critical, and some attributes are ones that just come down to "I'd
prefer if"
> 
> I don't pretend that I'm perfect at UI design, and I don't think that
> all preferences are bad. Thats not even what the essay says, it says
> that all preference additions must be carefully considered, and that
> addition of a preference does have a cost (contrary to what many
> believe). Does the theory of product selection attributes say that you
> should add preferences without considering the costs at all? Or how is
> it disagreeing?

I wouldn't call it "addition of a preference", which it would be if this
mode wouldn't have been there let alone default in previous versions of
the software. It's a question of "how compatible is the new Nautilus to
large parts of its existing user base". When talking about UIs, a lot is
based on what a user is used to do, on his or her background, not on
what would be ideal if he or she weren't exposed to any other artificial
(computer) user interfaces in the past.

> However, I find it interesting that you are the person who argues like
> this. How would you counter someone using your argument when he wants
> support for system V streams or some other (in you opinion) horribly
> ugly but used by important people feature in the Linux kernel. At some
> point you have to stick by what you (as a developer) think is best
> (considering input from other parties), instead of letting the
opinions
> of the person with most money decide.

I find it interesting that you bring this analogon when you are a person
who must deal with the differences between a user and a developer.
Besides, System V streams have never been in Linux so it's not something
where you would have to make an option for old-time users of it.
Developers can to a much larger extent be expected to adapt to the
environment where they're programming in than mere users of a computer
IMO.

> > There is a lot of good argument for UI that doesn't throw 1000
options at the
> > user, but the "remove everything" model requires that you know which
attributes
> > the majority of the user base consider in which light. Without doing
that
> > analysis of the userbase you don't know which attributes you can
remove.
> 
> No, and since you can't do such an analysis you have to choose by
other
> means. Things like experience, feeling, user feedback, what the
> competition do, research, etc. This is no exact science. We can't come
> up with the 50 optimial preferences according to some well defined
> measurement.

No one claimed that. But it is clear in my view that "one size fits all"
doesn't cut it in UI development, there has to be a "bandwidth of
preferences" people can _easily_ move in. This is not some power user
setting only 1% of users like (like e.g. "always on top" in a WM,
something where I don't have a problem if it's only accessible through
GConf). The default has changed and it is obviously non-obvious ;-) how
to revert to the old behaviour. This has been done as you admitted below
without billions of dollars of research that would back up the new mode
as so vastly superior to the old one that the old one can be dropped
like that (while this is not technically so, the user perceives it as
that). Now what is wrong with this picture?

> > It is also often about presentation of an attribute. Gnome for
example lets
> > me set the desktop background. To most users thats firmly an "I'd
prefer if"
> > thing - so why hasn't it been removed ? - because there is a sane
way to
> > let the user set it without throwing hard questions at them.
> 
> Yes, background settings, while not important for productivity, are
very
> easy to explain in a ui, and experience has shown that people find it
> very important to be able to personalize the look of their computer in
> basic ways.

Lack of being able to set a background hasn't put larger obstacles in
the way for people trying to get their work done. And I'm not making
this up, I actually know people working at locked down machines where
they just can't do it, they still get their work done. _Apparent_
removal of the former default way people got used to might be different,
but that's just my opinion ;-).

> > Browse v Spatial mode seems to be an attribute which is important to
users,
> > so arguing for removing it by simplifying the interface is actually
flawed
> > when you remove your head from the cardboard box of purist-UI and
look at
> > the real world. 
> 
> I'm not so sure that Browse vs Spatial is more important than the set
> desktop background feature. Don't you think we'd be mightily flamed if
> we remove that feature? Don't you think its important to users?

People would be upset, no question. But that's not what we're talking
about. The crucial point is that the former default behaviour is -- at
first sight -- gone, people have gotten used to it and they won't know
at first glance how to bring it back. As others have pointed out, this
is not perceived as a property of the home folder or something, but as a
setting of the file manager itself. Putting it as some kind of a
different program in the start menu doesn't cut it as well, because it
isn't perceived as a different program.

> I'm not arguing for removing browser mode. Its there, availible by
> default in both the start menu and context menu (and you can easily
drag
> it to your panel). However, I believe that for the majority of people
> who use a graphical filemanager spatial is better or as good as
browser
> mode, and for the times when you need browser mode its easily
> accessible. Since both modes are easily accessible I don't think its
> worth having a visible preference for which one to use as the default.

>From my days as an armchair UI designer I know that you should use the
principle of least surprise to the user which is clearly violated here
unless you only look at people who like the new default (and won't go
looking for something else) or who never have used file managers before
(because they don't know any other way). Both groups won't be in the
majority I guess.

I admit that I personally don't often use the file manager when being
alone. I always use it when dealing with normal users because this is
what they are accustomed to use. I know without asking her that my wife
will be negatively surprised by the new behaviour, even if she might (or
not) like it better once used to it. But that's all it boils down: The
are things you get used to and you need a pretty good argument to change
them and an even better one to make it so non-obvious how to bring it
back as has been done with the new Nautilus default mode. IMO the
arguments you bring for hiding the setting in a place where one wouldn't
expect it at first glance -- a setting that enough people will demand,
trust me -- don't fulfil the criterion of being good enough. Leaving it
that way will drive some people to using other file managers or even
desktop environments unless they are able to change it and the way to
change it is intuitive enough.

Someone -- maybe even I -- might be annoyed enough by the rigid "no
preferences at almost all costs" stance that seems to be predominant
with GNOME that a TweakUI-alike thing may get written to give choice
back to the people who want it, to give an alternative to "choice by
changing the application/environment used". Trust me, this isn't
something I would love to do, everything TweakUIish must be a hack, a
workaround to make settings a not minor people wants to have _obviously_
accessible.

> Now, this is my opinion of course, and I don't have billions of
dollars
> of research to back it up, but so are all design choices I make in
> Nautilus. There are lots of way a file manager can behave, and we
don't
> want them all in the preferences dialog, or all combinations of them
> described in the docs. To avoid this I try to err on the cautious
side,
> because once a preference is in its very very hard to get rid of.

It's even harder to get rid of people used to the old default, meaning
that is something I would try to avoid at some cost. 

> Where exactly would this be? A dialog that shows up when you switch
> mode? It sounds like a normal preference to me, except not in the
normal
> place for preferences.

Again you don't seem to put much value in not surprising people then --
either it's a preference or not (it is ;-). Is it a preference a large
number of people want to have? Surely, especially if you look at the
people who use Nautilus now. Is it major enough to warrant a preference
setting where users expect preferences to be set? Given the emotions in
this thread, I'd say it is major enough to be put in the preferences
dialog. Besides, preferences only make sense to be set "where
preferences are set for a specific object" (this is the preferences
dialog for an application), rather than to be hidden behind a purely
artificial "this is another application even though it uses the same
bits" analogon like putting it as a separate item in the start menu or
by putting the preference where it doesn't belong, i.e. in the settings
for the home folder.

Enough rambling for now. If I sounded too rude in some places, my
apologies.

Nils
-- 
     Nils Philippsen    /    Red Hat    /    nphilipp at redhat.com
"They that can give up essential liberty to obtain a little temporary
 safety deserve neither liberty nor safety."     -- B. Franklin, 1759
 PGP fingerprint:  C4A8 9474 5C4C ADE3 2B8F  656D 47D8 9B65 6951 3011





More information about the fedora-test-list mailing list