Prerequisites
 What you should know before reading:

• How to examine a text file
  Use cat or less to examine a file whose contents is ASCII text.
  cat lists the file to the screen and you can use Shift+PgUp (or Dn) to scroll the screen up and down to look at it (as long as it's a small file).
  less is a viewer program where you can use the cursor keys to move around. To get out, type the letter q.
  The file of interest to LILO hackers is lilo.conf, found in the etc directory.
  Example:  less /etc/lilo.conf

• How to edit a text file
  There are several commonly available editors, vim, joe, and pico.
  vim (an upgrade of vi) is the standard-fare (and full-featured) Unix editor. It's a usable choice when you can't find any other, and you can count on it being there regardless of which Linux you have. I figured out enough of it to be able to alter files, by reading the help that's included with it. :q gets you out of this one.
  joe is a full-screen editor that takes many of its commands from venerable old WordStar.  You can display an onscreen help menu by typing ^KH. ^KQ gets you out of this one.
  pico is even simpler than joe. Commands for this are shown right on the screen while you're using it.

• What the disk device names mean
  Beginners often confuse /dev/fd0 with /mnt/floppy. We can illustrate an important concept by explaining the difference.
  /dev/fd0 is a block device, accessed in a "raw" mode via its driver. This is characteristic of block devices found in /dev. These are made available by virtue of having had their driver compiled into the kernel or loaded (as a module) during startup. As such they do not need to be "mounted" and are directly accessible to programs such as dd and hexdump. They appear simply as numbered blocks of data to the corresponding program. So they are also difficult to manage.
  /mnt/floppy is quite another matter.  A file system is fabricated to bring some organization to this jumble of data blocks found on /dev/fd0. When you talk to the file system it makes the disk appear as though it has directories, files, and ordered blocks within each data file. This illusion is the work of the file system. The file system works by accessing /dev/fd0 in turn.  This is all brought together when you use mount--you tell mount to access /dev/fd0 and to present it to you as a nice file system which appears as /mnt/floppy. (Yes, you can change /mnt/floppy to appear as different names, and in different directories, and many people edit their fstab file to do this.)
  
  It's the same story for the hard drives. The first (IDE) partition is always found at /dev/hda1. This also is a block device accessed in raw mode. The very first block of hda1 is often used as a "boot sector"--not strictly necessary in Linux, but definitely required for DOS and others. The rest of the blocks in /dev/hda1 contain all your directories, allocation maps, and files in a fairly complex arrangement. To make sense of all that data, you install a file system (ext2 being the current favorite), which will access /dev/hda1. You mount it and give it a name--probably / , the root file system. Now you can see subdirectories and files.  If you have an hda2 (a second partition), you can mount it with another name (/usr, /home, or make one up, like /extra)--it will appear as more directories, but now you know that it is an illusion created by the file system that uses /dev/hda2. Naturally this applies to other disk types you may have; hdb1 is a second drive, sda1 is a SCSI drive, and so forth. The partitions don't have names. You assign them names with mount.
  
  What if you had an MS-DOS disk on hda3? The blocks in /dev/hda3 are just as valid, but ext2 isn't going to recognize anything there (in fact you will probably get a message suggesting that you've picked the wrong filesystem). Yet, if you tell mount to use msdos as the file system, then you can access it, directories and all! So, mount associates hda3 with (say) /mnt/C-DRIVE using the command mount -tmsdos /dev/hda3 /mnt/C-DRIVE.

• What is all this about disk partitions? How is hda different from hda1?
  Partitions are all about dividing up a disk into smaller, "logical" disks. The usual use for multiple partitions is to host multiple operating systems (this was the original use).  Nowadays it is common to find disks divided up because certain operating systems cannot access a drive larger than a certain size. Partitioning is accomplished using a partition table.

The partition table is located on the very first block of the disk drive.  The drive (using an IDE drive in this example) is found at /dev/hda. hda is the whole disk. If yours is a 4GB disk, then you will have 4GB worth of disk blocks accessable from /dev/hda. That very first block is special.  It is called the Master Boot Record. It contains a program (no larger than 446 bytes!) that determines how your system will boot. The last 64 bytes of this block hold the partition table.  Each entry (16 bytes) yields about 8 pieces of information. That's all it takes to describe a partition. Obviously there is a limit of four partitions using this scheme.

The information tells where each partition starts and ends. It is the responsibility of the program that sets these (fdisk, in at least two of the operating systems), to make sure they don't overlap. Once they are in place, the disk driver pays close attention to these limits and does not go beyond them. Thus, each partition can remain independent of the others.

The usual need to go beyond the original design has given rise to the extended partition. In this, one of the partitions is designated as extended. It is still assigned to a particular section of the drive, but now its contents can be further carved up--another partition table is set up to describe these logical partitions.

The partitions in the original scheme are called primary. They will be numbered as hda1, hda2, hda3, and hda4.  If one of them is used for an extended partition, then it's not available as primary. Because there can only be one extended partition, it is customary to use hda4 for it. The logical partitions are numbered starting with hda5. The program disk druid is known for its habit of setting up hda1 and then jumping directly to hda5 and higher for the rest. If you run fdisk or disk druid you can use the display function to examine the structure of your disk partitions. Also, the structure is shown when Linux starts up. Look in dmesg (or watch the bootup) for a message like this: 

Partition check:
 hda: hda1 hda2 < hda5 hda6 >

/dev/hda is used only for accessing the MBR and managing partition information.
/dev/hda1 and up can be used for an operating system, an additional drive, Linux swap, etc.
LILO can be used to start any of the operatings systems located in a legitimate place. "Legitimate" is a concept determined by each particular operating system.  To wit, MS-DOS
     won't start from an extended partition, but Linux will.
mount can mount the file oriented partitions as /, as /home, as /mnt/dos, etc.
Some partitions aren't file systems, such as the swap partition.

Each partition has a boot block as its first block, as mentioned before. Do not confuse these with the MBR.  The MBR is the first block of /dev/hda.  The boot blocks are the first block of /dev/hda1, /dev/hda2, and so on. This distinction is important when considering how to configure LILO.

There is much more to be said but it is getting away from lilo and is a subject probably just as large.  If there is enough interest I will create some pages for that.

For now, I will say only the following in regard to lilo:

If you mistakenly trash the MBR, or the boot sectors of various partitions, you can recover them provided you understand what's going on and have made the proper boot diskettes!

The lilo.conf line boot=/dev/xxx is what determines where LILO gets written to.