Re: SEUL: Re: Comments

[William T Wilson <fluffy@benatar.dunadan.com>]

On Mon, 5 Jan 1998, Star's End wrote:

> Why not let them hose there data? It will teach them not to just turn off 
> the computer. Even windows95 doesn't let you turn off the computer without
> going through a shutdown process.

Because the point of a usability-oriented distribution is to cut down on
this sort of thing.  One of the big advantages of SEUL over Win95 is that
it will be much stabler.  We want it to stay that way.

Probably the only fix to this will be when the kernel developers add
journaling to the filesystems.

Modern Win95's if you just turn them off will complain the next time you
boot them and then run scandisk.  Just like Linux does.  Which one is more
likely to lose data is open for debate.  It depends entirely on the
situation at the time of the power failure.

In general an unloaded system (as the state of a system when a user turns
it off without bothering to shut down) will never have any trouble.  A
system processing the mail queue on the other hand will suffer a most
grievous fate.

> It isn't the programmers mentality who is at fault here, but the millions
> like mom who jump on the bandwagan then scream when the ride becomes
> to bumpy for them.

Unfortunately, these are the people who we are targeting here.  Smoothing
out the bumps is the whole point.

> The answer to that is don't distribute them at all. If the licensing doesn't

Well this is hardly useful.  Unless you feel like writing an app up from
scratch, which I personally don't, it's easier to jump through some
licensing hoops - especially to maintain compatibility with a very common
standard program like the JDK, xanim, <asbestos> or qmail </asbestos> -
than to write our own.  There will be enough to do.

> Instead of bending over backwards to accomidate these people, simply cut them 
> off.

In the case of the JDK, Sun really doesn't care.  In the case of xanim,
there's nothing the author can do - Apple and Intel (I think) own the
copyrights and THEY don't care.  Qmail is just weird.  In the case of
qmail there are alternatives.  Not for xanim.  And kaffe despite its speed
is not ready for prime time. 

> > written by us could.  An installer that would log on to the network,
> > retrieve the applications, compile or patch them if necessary, and install
> > them.  All the user has to do is click yes on the license agreement.
>
> The owner of the license could also make sure that there product is 
> compatible with the standard linux file system layout and not try to
> break everything.

Would be nice, wouldn't it.  Unfortunately, not all things work this way.
There will be programs that require patching and which cannot be
distributed in their patched form.  JDK, qmail, xanim for example.  There
will be more.  If it isn't GPL it will probably have this problem.

> Applix did a very good job of making sure there product didn't cause problems.
> Why can't other companies. Especially those that don't distribute free
> software.

They can.  AcceleratedX is easy to install, Executor is easy to install,
Applix as you mentioned, OSS, a lot of things work this way.  Not all,
though.

> Such a system exists. Is called ezppp. Once configured, you choose where
> you want to connect to, and hit the connect button. It ain't easier then
> that.
> My beef is that diald is not supported on this. But you can't have everything
> I guess.

Ho-hum.  Can't win 'em all.  :)  That's the sort of thing we need.  Why
haven't I heard of it?

> Do you have any clue here. I had to debug a windows box which could make
> to different connection to the internet. Unfortunately neither would work, 
> and can't ever work, because the way windows does networking.

Usually, yes.  :)  What I'm saying is that Linux out of the box has IRQ's
tuned wrong, has bad modem init strings, often has the UART speed set
wrong, and that this confuses a multitude of new users and more than a few
experienced ones.  Windows' inability to have more than one internet
connection is just another Windows misfeature.  Let's not worry about it.
:)

> I'm beginning to suspect that you are a windows person who doesn't want
> to learn a different system. Its people like you that keep me employed. The

No, but I'm trying to see things from their perspectives.  And from the
fact that as of now the world is completely centered around Windows and
that everything you see is designed for it.  At least at first, we'll need
to be able to handle things like DUN scripts.  The SEUL project is for
people that don't want to learn a different system - at least a radically
different system.

> above script you mention is something I charge most companies an arm and a
> leg for.

Hate to put you out of business then... but I bet I could do it in a
couple hours with Perl, maybe more, depending on the complexity of the DUN
scripting language.  :)

> There is no such thing as devel versions of anything. If you have the

Sure there is.  I'm speaking of Red Hat here... you've got a package, and
a development package to go with it, many cases.  It's needed to compile
but not to run binaries of certain applications.  Disk space is cheap.
Leave the .h and .a files in.  That's all I'm saying.

> wrong library the install script should be able to detect it, but what
> do you do when you have two packages that use different incompatible versions
> of the same library. It happens more often then you think.

That's what RPM does.  We may end up using LD_LIBRARY_PRELOAD and
LD_LIBRARY_PATH, but the user should never have to see it.

> What you need is a staic environment.  That doesn't exist in linux. In
> fact most would say the chaos in the linux world is its biggest strength.

The chaos makes the development fast, and it also keeps the corporate
users away.  Best of both worlds... :)

> It just you. This has never happened to me. I have never heard of happening.
> If this is really problem, take the performance hit, and compile the cdrom
> driver in the kernel, and not as a module.

I do, actually.  All my kernels are monolithic at this point.  There's no
speed loss and maybe a couple hundred K of memory used, tops.  And all my
stuff always works.

> I think you mean window manager not file manager.

no, I mean file manager.  That's what My Computer or Explorer is.  Window
managers are just so much fluff.

> I agree here, but I don't see this issue going away. There is no standardized
> window maganer on unix, so again we go back to if you want an open system
> you have to put up with inconvience.

And this is one of the inconveniences we should do away with.  I would
suggest either FVWM95 because it's so similar to 95 or Enlightenment,
because it's so good.  Me personally I use FVWM rigged up to work like
MWM.  That IS just me.  :)

> I disagree. What we really need is people working on linux native apps.
> WABI is as good as it gets for windows emulators. But it is expensive.

WABI only runs 16 bit apps, and is no longer under development.  I'm
thinking more like SoftWindows 95.  Now THAT would be useful, but no one
will build a Linux version.  :\

> Intutitve is a strange word. For a person like me who has been doing
> everything in vi for ten years, the windows notepad is not intutive.
> You need to use a mouse, and the hot keys are all wrong. I mean who
> thinks of <alt>fx to quit a program. Its always been :q! to me. 

Yes, but it's much easier for you as a vi user to figure out notepad than
for a notepad user to figure out vi.  That leads me to conclude that
Notepad is easier than vi.  I doubt very many people would argue.

> The wizard idea didn't come from MS.

It's still a good idea, regardless of its origin.  And MS uses it
extensively.

> The answer has been given many times before. Make a distribution that is
> easy to install, complete, and preconfigured. That's all that windows is.

For the first time I'd be inclined to agree.  Windows is easy to use
because it works without any tweaking.  You can still tweak it and make it
better, but you don't have to.  Personally I don't find that there's
anything magical about a GUI, or that there's anything magical about point
and drool.  But that's what people expect now, so that's what has to be
delivered.  Regardless, there has to be some sort of cue to the user as to
how to proceed or at least how to get help.  To me, that's what makes
programs easy.  Finally, the program should not require any more knowledge
from the user other than what the user wants to do.

> Anyone who has had to fix anything that goes wrong on windows know exactly
> how hard and how cryptic everything is.

Not hard at all.  Just reinstall.  ;)

> Keep the user who doesn't want to how the computer works away from the
> part that are critical.

exactly.