When did Linux get so complicated?

Always expected and well documented.

> but that some sneaky bullshit was also overwriting part of unbound’s config using DHCP config data too.

Absolutely unacceptable and should never happen. I wish the author had actually shared details. That's a bug, and it should be reported, publicized and fixed.

> This is your weekly reminder that cosplaying as a sysadmin can waste your Friday evening and Saturday morning, and to add your own time into the cost/benefit calculations of running servers at home or on a cloud service.

This illustrates the value of OSes and OS projects that don't change gratuitously, that don't change to differentiate themselves because of their need to appeal to businesses, that don't make decisions about things based solely on usage and profits.

This is why I use NetBSD. The BSDs are clean and quite consistent.

If someone wants to pooh-pooh systems administration, they should at least direct their ire properly. Nobody's forcing you to use messy, inconsistent Linux distros :)

On the first page of results on DuckDuckGo for the query

  resolvconf unbound configuration

https://wiki.archlinux.org/title/Unbound#Using_openresolv

which describes the issue/feature and links to the upstream documentation. So that seems pretty Google-able.

You can also get it via

  man resolvconf.conf

To return to

> at what point...

Maybe about a decade, in the case of openresolv? Looks like it's been the most common resolvconf implementation since the early 2010s at least, maybe longer.

I could see myself running into this if I tried to run a DNS server with a dynamic IP for some reason (maybe I fix the IP with a DHCP reservation instead of configuring it on the server?). It would probably take me some time to figure out what was going on, and I'd be frustrated, too.

If you don't use all of that crap and stick to a simple workflow, things become easier to manage

In fact, your distribution is what makes it complicated, they ship with bloat you don't need

When our needs got so complicated.

We have dynamic networking environments now, where a static DNS resolver configuration just doesn't make sense any more. And when one DNS server is unavailable, we like to have the system remember that and not try it every single time because the resolv.conf model doesn't allow different process to share that information.

Mostly backwards compatibility works fine if you just replace /etc/resolv.conf with a static file.

But if you do that, then various modern expectations will fail, such as what happens to your DNS when you connect to a VPN, or use the wifi at your local coffee shop, or your LAN connection dies and your laptop automatically switches to some available guest wifi or perhaps your phone hotspot.

> Well, until you find some random post online where some poor person is being berated for daring to suggest a slightly hacky way to fix things, and someone with zero personal skills wades in with “that’s a dumb way, why aren’t you reading the docs and doing it correctly like this? You’ll break things”.

I guess that one reason might be that when I installed Slackware from floppies on a 486 there wasn’t that much stuff in that Linux system. Few things worked, but the way I remember it, the error messages were immediate and clear.

Nowadays things break suddenly without even error messages. Some times I do get some hint in one of the log files, often I get no clue. I’ve developed a conspiracy theory that people don’t want to be the ones that report an error because that makes their software look bad when the cause might be someone else’s code, so they just catch an exception or ignore an error code and go on, until so many things have failed that there are no workarounds left.

This is why I always look for the least "professional" way to do things. Instead of Linux, I'll use Windows. Instead of messing with BIND9 config files, I'll use something like TechnitiumDNS, which has GUI tools built-in. I lose flexibility, but usually don't need it.