> Now is this useful? The short answer is no. The long answer is also no. I can think of exactly zero uses of this experiment (and I must be pretty crazy for doing it).
Well, at the very least it's a very computationally expensive test that managed to catch a few packaging bugs.
In the bad old days we used to do the "kitchen sink" installs of Redhat because our machines weren't connected to the internet nor did they have access to the install media after install, so if we needed a package and it wasn't already installed it was a big problem.
Even today it's kind of a headache to install packages on machines with no Internet access, usually the easiest solution is to download everything and setting up a package repo on the LAN.
If you want real fun, try setting up a nontrivial node.js project without direct Internet access.
Which is definitely the reason to do it. Probably fleshes out most of the cornercases possible to experience. Further, HDD sizes are huge so just calling the whole enchilada install a standard install class makes a kind of sense. Only need to worry about picking and choosing until the whole enchilada becomes small enough to not be worth the decision process to minimize.
Well, at the very least it's a very computationally expensive test that managed to catch a few packaging bugs.