Hacker News Reader

Comment by badatnames | Hacker News Reader

Where "basic" follows a highly nuanced alternate definition, where if you step out of line without realising it, you lose 3 days debugging a complex undocumented interaction between 4 enterprise crapware daemons only to find out the answer is "you're wrong and won't fix" buried in a 6 year old bug report, when all you did was plug in a USB mouse or try to set the default audio output device, or something else.

Total and utter propaganda, it's underdeveloped steaming garbage

View on HN

I think the problem is that it's overdeveloped rather than underdeveloped.

You know it's underdeveloped because everything is described as "basic" or "simple". There's no such thing in any complex design, yet somehow for all his purported seniority, he keeps pumping out designs using the same description.

The reason for seeing overcomplexity in systemd components is usually that you never considered the use cases of other people or proper server management - mostly because in the days of olde, everyone cobbled together their shell script Jenga towers in seclusion.

If they don't have the problems that other people are having...maybe they don't need that complexity. Different use cases can use different solutions.

An Army unit crossing rocky and muddy terrain might want tracked vehicles and tanks. I don't need that in a vehicle I drive 10 minutes to the office in the city.

Analogies only go so far. Driving an actual tank with a cannon to the grocery store 10 minutes away is ludicrous, but my smartphone has gigabytes of both ram and storage and a CPU that runs at gigahertz frequency, but weighs about the same as a dumb phone that does far less. It's not the 1970's anymore. We can be inefficient and stop counting every single little bit where it doesn't count. Yes, where it counts, we still want to be efficient and mean and lean, like in transcoding, but for starting up services on a laptop with gobs of ram and CPU cycles to spare, the added expense of the tracked vehicle to the grocery store is what, an extra 100 milliseconds of loading data from nvme storage to RAM? Yeah, don't port kubernetes to a microcontroller. Focus on the interface you have with it and let the complexity just sit there, being harmless.

> Driving an actual tank with a cannon to the grocery store 10 minutes away is ludicrous

And here's me thinking it would be a lot of fun...

I think it’d really help on the highway with traffic.

I don’t need it all the time, but that doesn’t stop it from being useful.

This. I can see the utility of systemd and probably would have liked to have it when my team was riding herd on a DC full of Sun E15Ks and HP Superdomes. But for basically everything else that I use Linux for, every time someone says "systemd does this", at least 50% of the time I don't need that, and most of the rest I could do with with other tools with vastly less complexity.

Not "this". By that logic most people do not need backups too. Rebuilding working service state after config change is non-trivial, requires using admin-only kernel interfaces, and said kernel does not care about these types of problems. Meaning average user ends up simply rebooting rather than trying to figure why enabling wifi makes printer stop working (and such). We cannot have network stuff controlled by multiple programs, because linux kernel is broken like that. This thing is the userspace part of the monolith.

By that logic most people do not need backups too.

You need to go back for remedial logic.

Those jenga towers were terrible, many didn't work very well. Very few had good monitoring, ops, reliability, least-privleges.

It's unbelievable to me to hear people advocating for not having good broadly usable systems, to advocate that sysadmins really should spent their time and energy building bespoke custom cut weird poorly invested in little things to suit their own fancies. It's not helping the company to do this. And to me it's such a weird antisocial vibe, that is radically vocal, that shows up all the time to proclaim everyone everyone else is doing the wrong thing by working together and making tools that fit together, have cohesion. Theres never any suggestions for how or what that looks like either, just the image of not working with others is the only thing presented to sell itself. This is din. This is havoc. This is hell. Don't do it people. And alas, we have to endure this attitude every single time. The forces against cooperation are doggedly persistent and vocal. The forces for just build your own thing, this unreasonableness, is radically overrepresented.

This is why I was so hopeful this submission was going to be a nice new post kind of re-selling where we are today with systemd. Because the pro-social forces are weirdly not bothering to get the social / cultural victory. And those driving us apart telling us to each do our own thing, figure it all out yourself: they are endemic, dime a dozen on social media.

Actually I didn't say anything about overcomplexity and I wasn't thinking about systemd as an init system at all. That wild misinterpretation, which you knew to (partially) hedge with that "usually", is more or less the proof in the pudding frankly.

> The reason for seeing overcomplexity in systemd components is usually that you never considered the use cases of other people or proper server management

What use cases of "proper server management" does fucking with my resolv.conf(5) solve?

If I have a laptop I can perhaps see updating values per DHCP or IPv6 RAs, but once I set up a server up why the hell is it a link to /run? (I once had to use chattr +i because of brain dead "cleverness".)

In over thirty years of running Unix-y systems I never needed my resolv.conf "managed", so I have no idea why Ubuntu (for one) defaults to fiddling with it.

> What use cases of "proper server management" does fucking with my resolv.conf(5) solve?

For one, it avoids telling all processes and their mothers the IP of the DNS resolvers they should query, but hands them 127.53 instead. If you then update your resolvers using systemd-resolved, for example because the network changes to a failover line, or because puppet pushes an updated catalog, or you want split DNS for a VPN (like, send *.corp.local to the internal DNS server or something), that is all pretty easy. There are more use cases on servers, but of course you most obviously benefit from this on laptops that roam between WIFIs with different upstream DNS server config.

Just one thing: why that resolver work *need to be" in systemd package ? Why we can't have few implementations, as it always have been, that do different jobs, with overlapping funcionalities ? Why we ever need some gigantic codebase guarded by corporations that we can't easily replace ? Codebase that is uncooperative with other implemantations. And developers.

And moust importantly: why they make everything more stupid and windows-like trash ?

And I assure you: what you described as "easy" is not true. I is just few basic usecases they baked in. And when you neeed something a bit more specific in some direction you are not allowed to have that.

Systemd is a trash, atm.

> For one, it avoids telling all processes and their mothers the IP of the DNS resolvers they should query, but hands them 127.53 instead.

What "all process" are you referring to? What kind of software on Unix-y systems is not using gethostbyname(3) (or similiar)?

(And the context of this sub-thread is "server management", which I'm not sure how often has VPN running on it and doing things like split (horizon) DNS.)

You should install an actual local resolver for this, not systemd.

Reply on news.ycombinator.com