Why is sysinstall considered end-of-life?

[Ted Mittelstaedt]

----- Original Message ----- 
From: "Tore Lund" <toreld at netscape.net>
To: <questions at freebsd.org>
Sent: Sunday, January 07, 2007 3:45 PM
Subject: Re: Why is sysinstall considered end-of-life?


> Robert Huff wrote:
> > (Personally, I think there are also points where the correct user
> > behavior is not intuitively obvious.)
>
> An understatement.  There are situations where sysinstall is positively
> quixotic.  I don't mind the simple character-based interface.  But I do
> find it worrying that I sometimes cannot know what sysinstall will do
> next.  In any case, this is bad publicity for FreeBSD since sysinstall
> is the first bit of FreeBSD they encounter.
>

All of this is true.

> Time and again we hear rumors about a new installation program.  Is it
> actually nearing completion?  Keep in mind that many of us do not even
> consider getting involved as long as we believe a better program is
> under way.

There is no new installation program underway.

This comes up every year or so on the various discussion lists, everyone
bashes sysinstall and claims it makes FreeBSD look bad and when are
we going to get a replacement, etc.  The arguments die away when faced
with the following cold realities:

1) You can probably get consensus from everyone that sysinstall is ugly
and needs replacement.  But your never going to get any consensus on
what the replaement should look like.  And any replacement is going to
have places where the user cannot know what it's going to do next, that
is just the nature of install programs - it is due to the fact that
different people
interpret things differently.  What is obvious to you isn't obvious to
someone
else.  And, when is the install program going to cross the line between
acting as a install program and acting as a training video?

Review the steps needed to install a self-signed SSL certificate into
Microsoft Internet Explorer 7, and then come back and tell me that
those steps are more intuitive than sysinstall.  Yeah, right.  Face the
facts, boys.  Every year, computers get more complex to operate, and
every year, the Average User is paying more and more to have a tech
set the computer up for them.  Open your eyes and look around.  People
think nothing of paying $30 to have a tech install Microsoft Office on their
new Windows PC for God's sake.

Who really is sysinstall's audience?  The average l-user?  Or the average
technician?  If it's the average tech, then who the hell cares how ugly
sysinstall is?  You think sysinstall is bad, you ought to see the diagnostic
interface
the average auto mechanic has to use to troubleshoot your car.  If you are
not the ultimate end-user for the FreeBSD system your installing, then
you don't have any moral ground to make a call for pussifying the FreeBSD
install program.  I can tell you that for myself, every FreeBSD system I've
installed in the last year and a half has been for OTHERS to use, NOT ME.

2) There's an immense amount of effort that has gone into sysinstall and
it's libraries.  Your talking about taking on an old, established program
that
is pretty throughly debugged, a program that is like an octopus in the
amount of icky, ugly mucking around with config files and such that it does,
and replacing this with a new program that is going to have all of the
intelligence and institutional knowledge in it that the old program does.
And furthermore if this replacement is to ever get traction among the
userbase it's going to have to work PERFECTLY in the FIRST version
that is released, otherwise everyone is just going to turn their back on it
and keep using the existing sysinstall.

3) The largest complaint about sysinstall is that it's not graphical.  The
problem is that a graphical installation program has some -severe-
constraints on it.  First, it has to work in ALL instances.  That means,
640x480x16 colors VGA screen.  You have a lot of people out there
installing on systems that have, for example, monitors with inadequate
horizontal/vertical frequency ranges and very capabable video cards,
unless you force the X-server to use the original VGA resolution, it's going
to overdrive those monitors and the user is going to see a black screen
when the installation program comes up.  And the only way FreeBSD
is going to get a graphical anything is by using Xorg, and FreeBSD does
not maintain that distribution - so we are now dependent on the Xorg
group writing their code with no bugs for our installation program to work.

4) Installation programs by and large are not "fun" programs to work
on.  Most developers avoid them.  They are thankless tasks - you
don't hear squat for thanks from anyone when they work, but you make
the least mistake and everyone is on your neck.

5) Finally, sysinstall is a one-shot program.  You use it once, the system
is
installed, and you never have to touch it again.  There's lots of other
things
in FreeBSD that are critical things that will stop an installation cold.
Such
as lack of device support for some new piece of hardware.  These things
are much higher on the priority list than replacing sysinstall, a working
program.

Ted