It's been around two decades since I last did an LFS install, but I still find the knowledge I gained to be useful to understanding how boot loaders, the kernel, init systems, etc. work today—not just bare metal and VMs, but also in containerized environments.
Same, I'm closing in on twenty years since I did it on a Pentium 2 (which took a very long time to compile). But in a matter of weeks I had my own linux router that I built from scratch.
I was too much of a novice to completely understand everything I did but over time the puzzle pieces fit together and I realized it was just about the most solid foundation I could have engaged in. I'd suggest it to anyone that interacts with Linux and the time requirement to compile everything is far, far less than it was years back.
I spent a couple weeks planning a LFS install about 20 years ago. I then didn't have time to look at it all for about year, so I never went through with it. But I still learned a lot just documenting all steps and preparing with the worse case of not having decent internet if something went wrong.
Same. Any time I'm asked a question like "How can I better understand Linux?" by someone new, I tell the person asking to do LFS and to take the time to understand it. Even some long-time system admins and software devs would benefit greatly.
Building LFS was probably the best single learning experience I've ever had! I built it in highschool and got credit for an independent learning project. Taught me a lot of things that turned out to be useful in finding myself a career.
LFS, Crux, and Gentoo probably changed my life in profound ways 20 years ago. At the time, I thought I was kicking the tires on desktop Linux.
What I was actually doing was learning how to apply critical thinking in the real world. Every abstraction layer removed from a subsystem tends to require increased reasoning from the user. Maybe even a little ingenuity.
Same for me, I think I first tried Gentoo. Becoming completely obsessed how to make my system as fast and stable as possible I did the LFS tutorial. Before all that I had been using SuSE and was so frustrated how long it took to boot (~1 minute) and how cumbersome it was to build from source. On Gentoo/LFS that was just a breeze.
In a quest to better understand what was going on under the hood, I built an LFS a couple years ago, then a Gentoo after that.
I'd been a Ubuntu/Mint user until then, but after spending hours knee deep in Arch's superb documentation to explain the tools I was building for LFS I've become an Arch/Manjaro man (and Debian when I'm provisioning a new VPS for something that needs LTS)
I learned so much about what makes a distro a distro from LFS.
I'd reeeeeeeally an Alpine style LFS fork that is fully built with LLVM and muslC, but we're not quite there yet.
There's a current/working m68k Linux kernel? Cool.
It would definitely be interesting to try a machine that was contemporary with the earliest Linux releases. And apparently it can run on a beefier qemu-emulated virtual machine as well.
On the BSD side, I think NetBSD also supports m68k and vintage Macs, though I imagine something like a modern web browser would be too slow and heavyweight to run properly.
If you run LFS long time keeping it up to date, won't it get cluttered because of incremental 'make install' or its not meant to be run as normal desktop distro? Honest question, last touched it like 15 years ago.
Does anyone know if anything similar exists for Windows? Of course it would be hard due to the closed nature of it, but a similar in-deth look into Windows internals would be much appreciated for a UNIX guy like me.
Installing LFS is a ritual, a manifestation of the belief in free software. In order to affirm allegiance to the corporate world, installing Windows 11 with all telemetry enabled and an online account might be more appropriate.
Not as a hands-on project I don't think, for the reasons you listed, but there are books that do explain these things, and are probably the closest to what you want.
