Here's a project I've been thinking about for a good long time, and which I've finally decided to get under way in public: I'd like to try and build my own custom Linux distribution.

There's a whole slew of reasons I want to do this.  Obviously, I'd like to learn more about Linux from the inside out, which is never a bad thing.  I also want to get a flavor for the issues involved in working with Linux on that level -- why certain decisions are made about things, what the consequences are for making such decisions, and so on.  And finally, it sounds like a heck of a lot of fun -- the kind of tinkering that I could really sink my teeth into.

The first thing I want to nail down are the design goals for this distribution:

1. Lightweight.  I don't want to build anything huge -- just something that can boot from a CD or flash drive, run a few basic commonly used applications, and be able to save the user's settings persistently.  This isn't the same thing as trying to figure out how to cram the most applications into the smallest space.  Installing to a hard drive also is optional at this point, but I'd like to have that if I can.

2. Logical.  After reading about the GoboLinux directory structure, for instance, I thought I might try and implement something similar myself.  The existing nest of directories in Linux always seemed terribly counterintuitive and I wanted to take a step away from that if I could.

3. Useful.  The more real work I can get done with the distribution when I'm finished with it, the better.  This means I'll be adding in builds of applications like Firefox, Thunderbird, OpenOffice, etc., as I move along.

4. Well-behaved.  I'd like to try and make sure that this distribution plays nicely with things like ACPI, at least on the machines I run it on myself.  I don't plan to run using terribly exotic hardware, so a number of things (like PC Card support) can probably be omitted entirely from the kernel, and thus I can probably pare down some of the complexity that might come from trying to run everywhere.

5. Public.  Meaning that each step I take is going to be documented here -- not just when things go wrong (as I'm sure they will at some point), but when things go right, and why.

Obviously one of the first resources I plan to make use of is Linux from Scratch, with its instructions for assembling and building a distribution from the ground up, and Gentoo Linux, the other major wellspring for Linux do-it-yourselfers.  I doubt I'll use Gentoo except maybe in a compare-and-contrast sort of way -- at a glance, LFS's materials seem much easier to wrap my mind around.

Again, at first, I won't be doing more than gathering information and soliciting feedback at first, and discussing my rationalizations for specific choices as I go -- but that part alone ought to be enough to keep me busy for a while.  Wish me luck.