The luxury of understanding the problem (Shallow Thoughts)

The luxury of understanding the problem

On a Linux list, someone was having trouble with wireless networking, and someone else said he'd had a similar problem and solved it by reinstalling Kubuntu from scratch. Another poster then criticised him for that: "if the answer is reinstall, you might as well downgrade to Windows.", and later added, "if "we should understand a problem, and *then* choose a remedy to match."

As someone who spends quite a lot of time trying to track down root causes of problems so that I can come up with a fix that doesn't involve reinstalling, I thought that was unfair. Here is how I replied on the list (or you can go straight to the mailing list version):

I'm a big fan of understanding the root cause of a problem and solving it on that basis. Because I am, I waste many days chasing down problems that ought to "just work", and probably would "just work" if I gave in and installed a bone stock Ubuntu Gnome desktop with no customizations. Modern Linux distros (except maybe Gentoo) are written with the assumption that you aren't going to change anything -- so reverting to the original (reinstalling) will often fix a problem.

Understanding this stuff *shouldn't* take days of wasted time -- but it does, because none of this crap has decent documentation. With a lot of the underlying processes in Linux -- networking, fonts, sound, external storage -- there are plenty of "Click on the System Settings menu, then click on ... here's a screenshot" howtos, but not much "Then the foo daemon runs the /etc/acpi/bar.sh script, which calls ifconfig with these arguments". Mostly you have to reverse-engineer it by running experiments, or read the source code.

Sometimes I wonder why I bother. It may be sort of obsessive-compulsive disorder, but I guess it's better than washing my hands 'til they bleed, or hoarding 100 cats. At least I end up with a nice customized system and more knowledge about how Linux works. And no cat food expenses.

But don't get on someone's case because he doesn't have days to waste chasing down deep understanding of a system problem. If you're going to get on someone's case, go after the people who write these systems and then don't document how they actually work, so people could debug them.