Re: FW: Nautilus and Setup Tools

[Seth Nickell <snickell stanford edu>]

I'm not really referring to the technical implementation. No, We are not
developing a GNOME kernel. And yes! We should share everything we can
with other systems such as Swing, KDE, Mozilla, whatever so the desktop
is cohesive. None of these things have bearing on what I am saying.

> What I mean by operating system is the whole deal, including GNOME.
> However, to users, this "whole deal" is labeled Red Hat or Solaris or
> whatever, most likely. Because that's what it says on the box when you
> buy it.

Which is part of the problem, but not one I am able to solve. ;-)
Seriously, I don't care about naming games much (lest this start to
sound like Linux vs. GNU/Linux or something). I don't really care what
users call the desktop environment, whether they call it RedHat or
Solaris or GNOME. I don't care who the "credit" goes to or who users
think is making the "product". The point is 

For a large class of users, the desktop environment is the most
fundamental object they interact with, the desktop environment is what
holds their operating system together. When the graphical environment
fails to provide a good experience for the user, I think the final
burden rests with the desktop (perhaps the distribution can share this
as well, but as a desktop developer I choose to treat this as my
responsibility as well). If there's something missing that's needed, or
whatever, the desktop is the primary mediator of the user experience.
That's all I'm saying.

> If you want to extend into trying to solve all the UI problems of the
> whole operating system, instead of just creating a good desktop
> component, then you're going to have to look at working with the whole
> OS and various relevant open source projects, instead of just the
> GNOME Project. We have got to think bigger-picture if we want to have
> a bigger impact.

Right! We agree here, we need to be looking at the bigger picture of the
UI problems users are encountering. One of them happens to be system
configuration without relying on command-line tools, so I hope we do
something to address our half of that issue (the interface part), and
encourage the responsible parties (distributions, driver developers,
whatever) to implement the other half (the backend that interfaces with
a particular piece of hardware or distribution convention).

> Which isn't to say we shouldn't add XST, I'm undecided on that right
> this minute. (It's probably a 2.2 question anyhow.) But it is to say
> that we shouldn't think of GNOME as an operating system, because it
> isn't. It's one project that makes up the set of free software
> projects that typically make up an OS.

I think a technical conception of the heirarchy of the system isn't
helpful in this case. I conceive of the system as being structured
around fundamental units such as libc, the kernel, and branching out to
things such X-Windows, to GNOME libraries, to GNOME, and to applications
at the very end. So as a developer I see the kernel / libc as the
fundamental building block of the operating system, since that's the
thing that depends on the fewest other things.

I think the majority user conception is very different. To most computer
users (note I do *not* mean Linux users, I mean computer users at large)
the "desktop" is the Operating System, it is the most fundamental unit
of the system to them. Its what everything else is built off of, because
its the thing that visibly mediates their environment. I think this is
true even of most "advanced" users on systems such as Windows and MacOS,
even if they intellectually realize that there is a "kernel" and various
libraries. If we would cater to this class of users in the future it
would do us well to at least consider how they might view GNOME. I think
they would consider GNOME to be the "operating system", whether they
know its called GNOME or not. 

Here's a thought experiment that I think validates this point: I think
if you let most of these users loose on a GNOME desktop running Debian,
or a GNOME desktop running RedHat, or even a GNOME desktop running
Solaris, they would identify these as the same operating system. On the
other hand, I think if you showed them a RedHat desktop running KDE, or
a RedHat desktop running WindowMaker and a RedHat desktop running GNOME
they would not necessarily identify them as the same operating system.
Perhaps the branding cues would be enough to tie the two together, but
remember that the branding cues are artifical in some sense, you are
introducing them to your concept of the operating system. I think most
users asssociate the operating system with how it looks, feels and
interacts rather than what kernel it is running or the packaging format.
Actually, I think this is part of what results in confusion when I
explain the concept of a distribution to people with an interest in
Linux.

Because many prospective users will view the GNOME environment as the
most fundamental level of the system I think GNOME developers would do
well to adopt a "buck stops here" attitude. When there is a UI problem
with the system it *IS* our problem, even if its because the GNOME
interface doesn't cover a particular area yet (such as system
configuration). Maybe we don't always have the option of fixing it (for
example, the filesystem tree isn't laid out optimally for a user
desktop), but we should considering the options. Because to many users
*we* are the most fundamental unit of the system, not the kernel, not
X-Windows. The user may not know they're running GNOME, but GNOME is the
primary force shaping their environment (assuming they are running
gnome, of course ;-).

-Seth