new sysinstall

Garance A Drosihn drosih at
Sun Aug 24 00:18:50 PDT 2003

At 2:25 PM -0500 8/23/03, leimy2k at xxxxxxx wrote:
On Saturday, August 23, 2003, Matthew Dillon wrote:

    The biggest problem with any install occurs when
    you go through all the work of installing the system
    onto your disks, reboot, and... nothing happens.

That is an interesting point but don't we already kind of
have a running system when we first boot a CD?  Doesn't the
kernel on the CD already "work" if we make it to the installer?
If so... what is the purpose of installing a really small
installation to install the full installation?
Booting from the CD is not a *usable* system, it is only a
bootable system -- and one that is on a read-only device.
During my first freebsd install, I was overly impressed by
all the packages available to me.  I spent at least an hour
reading the one-line descriptions of many packages.  I
selected maybe twenty or thirty, and learned about many others
that I had never even heard of before. Very interesting, but
very time-consuming.
Much later in that very same install, I went to configure X11.
Something went wrong there, and I ended up with the display in
some totally useless state.  I couldn't see a thing on it, and
had no idea how to get back to a state where I could see what
was going on.  I had to ctrl-alt-delete out of that state.
When the system rebooted, the partially-installed system was
totally unusable.  I had to start back at formatting the disks.
Yes, this could be laughed off as a "learning experience", but
I see no redeeming value in the hours which were wasted.  If
anyone had wandered into my office at the time, and laughed at
my "novice mistake", they would have been met with some very
serious hostility.
So now I install a minimal system.  The only ports I add are
bash and cvsup-without-gui.  I set the password for root,
create my regular (non-root) user account, and reboot.  I have
a usable system.  I can then add packages, or compile ports,
or configure X11 (which I hate with a passion).  But if
something goes wrong, I will most likely be able to reboot
into a still usable-system.  I also might have the option of
ssh-ing into the machine, and fix whatever's wrong without
doing a forced reboot.  Also, if something does go wrong,
I'll know *exactly* where it went wrong, and it will be much
easier to have logfiles with info that will be useful for
debugging the problems.
I certainly like the direction that Matt described.  Users
may still have to start over if they find out that they
made a serious mistake with partition sizes, but for all
other mistakes they won't have to start over from scratch.
Garance Alistair Drosehn            =   gad at xxxxxxxxxxxxxxxxxxxx
Senior Systems Programmer           or  gad at xxxxxxxxxxx
Rensselaer Polytechnic Institute    or  drosih at xxxxxxx

More information about the Kernel mailing list