This Gentoo simplified guide assumes familiarity with Linux and
certain aspects of system administration. In particular, it assumes
you've already made up your mind about several possible options for
the installation process and future life of the system, such as
partitioning, file system choice etc. It also assumes you want a
custom-made OS from start, so it uses the stage1 files to minimize
wasted effort and old, almost unused packages filling up your system
partition. 

Knoppix boot (working with a chroot'ed shell window and with a host
shell window)

- boot Knoppix and configure Internet access

- set up partitions and make the root filesystem

- mount root partition (in this document, the example mount point is /g)

- mount proc filesystem
  mkdir /g/proc; mount -t proc none /g/proc

- install stage1 (get it from the Internet somehow, or from a Gentoo CD)
  cd /g; tar -xvjpf stage?-*.tar.bz2
 
- configure /g/etc/make.conf -> USE, CHOST etc (and why not choose your
  nearest mirror now anyway, GENTOO_MIRRORS, SYNC...). Doing it on the
  host system allows you to use emacs:-)

- enter into chroot'ed ennvironment
  chroot /g /bin/bash
  env-update; source /etc/profile

- get a portage snapshot, two options:
  emerge sync (also emerge-webrsync)
  or get a portage tarball from the mirrors and extract it in /usr

- get the files for bootstrapping the system
  If you want 2.6, then use scripts/bootstrap-2.6.sh. You'll probably
  want udev then, so pay attention at the udev section later on.
  cd /usr/portage; scripts/bootstrap.sh -f
  After the first couple of files are down, the next step could be
  parallelized on another chroot'ed shell, greatly improving the
  installation time on a slow link. With a nearby server (LAS, for
  instance), you can skip the parallelization approach.

- bootstrap the system (i.e. get a minimum set of tools for compiling
  software packages)
  cd /usr/portage; scripts/bootstrap[-2.6].sh (takes quite some time...)

- build a base system
  Humm... since I decided to include nptl from start, I emerged first
  gcc and glib (I think), so as to start with the right-hand foot:-)
  emerge -p system -> to see what will be done
  emerge -f system -> this starts to download needed files
  After the first couple of files are down, the next step could be
  parallelized on another chroot'ed shell...
  emerge system -> to do it
  Problems with emerging some piece of software:
	   - find out which package was the culprit
	   - "emerge -pv package" to "pretend" the installation and to
	     print verbose information about it: this gives which
	     arguments from the USE variable that the package is going
	     to use. One of the "+" options (those that are being
	     included in the compilation) is the probable culprit.
	   - soooo, try the emerge with "USE=-culprit_option emerge
	     package". Cycle through the possible options till you
	     find out a way of emerging the needed package.
	   - this approach is the first to try whenever an emerge
	     action breaks, at any time during system operation...

- get a kernel (at least 2.6, of course:-) from www.kernel.org,
  configure it, build it and install it
  cd /usr/src/linux-?/; make menuconfig
  make
  make modules_install
  cp arch/i386/boot/bzImage /boot/kernel-?
  cp System.map /boot/System.map-?
  cp .config safe_place (save your config somewhere else)

- configure /etc/fstab
  root, swap, cdrom, proc, sys and local partitions

- configure LILO or GRUB (in this example, LILO)
  edit /etc/lilo.conf (from lilo.conf.example...)
  /sbin/lilo

- emerge filesystem tools (required for the checking fs phase; in this
  example, jfs)
  emerge jfsutils

- emerge hotplug udev, be sure to follow the udev guide from Gentoo's site

- give root a passwd ("passdw" while inside the chroot environment)

- reboot into new system

- [extra info for udev only system] If the kernel complains about not
  being able to mount /dev/hda? (the root fs), no such device or
  something like that: I solved it by just booting the system with
  udev disabled (issue a "gentoo=noudev" as an option to the kernel
  comand line on GRUB/Lilo). This makes some sense, I just don't know
  what it is:-). My system did find the /dev/hda? root device with no
  problems with noudev, created some device file somewhere in /dev and
  on the next booting attempt without the safe "noudev" option, the
  kernel found the needed device, automagically:-)

- configure/assemble everything else from the native Gentoo system
  just booted. You now have a system to start with by booting into it,
  however you can go on emerging things in a chroot'ed environment
  based on a working system, such as knoppix (live cd), so that you
  can continue to use the machine while building a new OS.