RE: Open Source Supported Graphics Cards
On Friday, August 11, 2006 6:23 PM -0500, Paul Johnson wrote:
> On Friday 11 August 2006 14:41, Seth Goodman wrote:
> > On Tuesday, August 08, 2006 6:43 PM -0500, Paul Johnson wrote:
> > > On Tuesday 08 August 2006 10:38, Seth Goodman wrote:
> > > > Since the end-users we need to interest, if we are ever to
> > > > break out of the expert niche, will run X and use GUI's for
> > > > everything, being limited to low-end 2D performance will be
> > > > an ongoing problem.
> > >
> > > I thought the niche Debian was trying to fill was rock solid
> > > stability and reliability in a 100% free software format. If
> > > I'm confused, let me know.
> > <OT discussion>
> > That's a reasonable goal, even a good goal, if you are willing to
> > remain a small, exclusive club. If you believe that people who
> > use Debian need to be comfortable with the command line, consider
> > natural language as a second language behind PERL and be fluent
> > in regexp's, then it will remain a terrific operating system for
> > the few. Maybe this is what most people in Debian want. I'm
> > relatively new here, so if that's the case, please educate me.
> It's not that hard if you use a desktop environment and use the
> desktop environment task during installation. Getting it installed
> is the tricky part, but you'll only have to do it once. And if you
> don't like aptitude, there's kpackage, and I'm sure there's Gnome
> frontends, and even a web frontend (if you're really brave or on a
> trusted network).
Me thinks you misunderstand. I didn't have any real problem installing
a Debian server or desktop. OTOH, I don't panic when asked to grep for
patterns, write a PERL script or (at least in the distant past) write
SED scripts. However, casual computer users cannot and will not be able
to do any of those things. Getting the desktop installed is only a
small part of the battle for a typical Windows user moving to Linux.
That step is probably the easiest for a computer noob, and the problems
will start soon after.
> > However, if you have a desire to bring quality, free software to
> > a wider audience, you're not likely to get there with the present
> > vision. For the majority of casual computer users, who are
> > hostage to a certain evil corporation, the GUI is not just a
> > convenience to be used after fully mastering command line
> > operation.
> Though if you were read the HTML installation manual, or even just
> the mastheads, you probably would have gotten a base install with
> KDE installed without much problem.
Of course I read the manuals before I did my first install. I'd give
them a B+ for experienced computer users, and a D for casual computer
users. They refer to all manner of things of which the casual user
hasn't the faintest idea, and of the large number of concepts they don't
understand, they are at a complete loss to figure out which are
relevant. For example, we all understand what a kernel is, what it does
and when you need to think about it, which isn't often. This is not
realistic for the casual computer user. Frankly, even if we did
successfully explain this in plain speak, I have no illusions that a
casual user could manage to build a kernel to run on their non-compliant
hardware. It's just not a reasonable expectation. For the experienced
user, it's just another task, and any time spent refreshing what you've
forgotten is time well-spent. In the Windows environment, hardware
detection and driver installation is largely automatic. Knoppix
approaches this level of hardware awareness, but Debian seems to lag in
> > We presently _require_ people who use Debian to do this, or they
> > are effectively hamstrung once it's installed.
> Only if you aren't reading your monitor during installation is this
> a problem.
That only gets you to the end of installation. Besides, the average
Windows user is not going to notice when hardware detection fails or
there is a broken dependency because of the hundreds of lines of, to
them, gibberish that scrolls by on the screen. We can watch this, they
Post installation, common tasks are not easily explained, and the
documentation is often inconsistent or downright misleading. That's
acceptable for experienced users. We have a sense when something
doesn't sound right and will look elsewhere. When something works
differently from the documentation, it's a challenge, not a brick wall.
It all depends on your experience and point of view. I'm arguing to
consider the point of view of would-be Windows defectors.
> > Why do we require this? It's not for technical reasons, but
> > because we believe it is _better_ for them as computer users.
> Hypothesis not supported by evidence present. Sounds more like
> pilot error.
I humbly disagree. And that attitude will hardly attract Windows users.
Deny a problem exists, and if there is something wrong, it's the user.
Yes, Windows users, by virtue of not understanding the insides of their
PC's, do commit an astonishing number of ID10T errors. The fact that so
many users of this class can successfully configure a Windows PC is a
testament to the thought that went into the configuration scripts. It's
still garbage code, but they did do some things right, and we shouldn't
be afraid to learn from that.
Here are a couple of cases for things that casual users can manage in
Windows PC's but would have great difficulty in Debian. The following
is not meant to say that Windows is good. It's not: it's crap. But
they did do some things right, and we ought to take notice.
1) Doing a backup. Windows includes a rudimentary backup utility, and
most PC's you buy today come with giveaway commercial backup software
preinstalled that is quite serviceable. You are generally led through a
wizard that sets things up properly for most cases. If you want to do
this manually, you will do a very well by backing up only the
"C:\Documents and Settings\" directory tree. What do you backup in a
Linux system, and how? A bare metal backup is not so bad, though you
probably have to search out and install the software. You won't find a
good explanation of what to back up in Debian because that's tricky
business, even if you are experienced. User home directories are easy,
but config files are scattered all over the file system. Unix, and
Linux that followed it, were not built for people who use the computer
as an appliance, with little knowledge of its internals. Windows takes
care of such users quite well, but it's commercial software and it's
very badly written. There's no reason Linux can't provide the same user
experience for those who need it, but something has prevented Debian
from doing this.
2) Configuring a home network. On a Windows PC, the user is guided by a
wizard to provide workgroup name, and they are given explanation during
the process of what it means to share information indiscriminately vs.
requiring credentials and, for advanced users (!!!), setting up
read/write permissions. This may seem silly, but the result is that a
total noob can get a Windows PC talking to on a local network and
sharing printers easily. In Linux, the user has to install Samba, and
then configure it manually with a text editor. The casual user is
befuddled by the fact that they should need a Windows networking server
at all (can't all computers just talk to each other?). They would have
no idea they even need to add a package, nor do they know what it means
to add a package from a repository, nor would they figure out what
package to add of the thousands available. Once someone tells them to
add and configure Samba through Synaptic Package Manager (forget about
apt or aptitude, they run from the command line, and BTW, what's a
dependency?), they now approach the Samba manual. Here, the user is
exposed to concepts like emulating a Windows domain controller, which is
total gibberish to them. We know that it is safe to ignore anything to
do with that in a simple network, but to the casual user, the plethora
of terms like this piles up quickly and becomes a brick wall. If you
want to share printers in Linux, you probably will deal with CUPS. Lots
of people who know what they're doing have trouble getting CUPS to work,
and I don't think it reasonable to expect the computer-as-appliance
crowd will be able to manage it.
3) Setting the time from a remote time server. In the Windows
environment, you go to a freeware site, pick a time server client,
download it and double click on what you downloaded. It installs and it
works, though the user doesn't have a clue what is going on. In Linux,
you start out by reading several manuals on the NTP protocol site. This
is a terrific resource, and I really had a blast filling in gaps in my
knowledge about the hierarchy of time servers and the methods for using
multiple time servers at different tiers to improve accuracy. To the
computer noob, who doesn't understand that there are delays in
communications, this would be pure pain and largely incomprehensible. I
doubt they'd get through the documentation on the NTP site and I further
doubt they'd manage to get the NTP daemon running, even if they do
figure out how to run vi or vim.
If you want more examples, you can pick almost any common operation and
it is substantially harder for a computer noob in Linux than in Windows.
That's largely because the folks who put together Windows, even though
they can't seem to write ten lines of code without doing something
foolish, appreciated the need to provide wizards that anticipate what
most people want to do and the level of computer knowledge they don't
have. On this mailing list, giving a straightforward answer to a simple
question, rather than a polite exhortation to RTFM, has been referred to
as "breast feeding" the noobs.
That's the nut Debian has to crack, IMHO, if it wants to see wide
adoption. Expecting unsophisticated users to decide in large numbers
that it is in their best interest to educate themselves on the internal
workings of an operating system is wishful thinking. Though the
sentiment comes from a good place, it is completely unrealistic.