New systemd Update Will Bring Windows' Infamous Blue Screen of Death To Linux (arstechnica.com) 154
An anonymous reader quotes a report from Ars Technica: Windows' infamous "Blue Screen of Death" is a bit of a punchline. People have made a hobby of spotting them out in the wild, and in some circles, they remain a byword for the supposed flakiness and instability of PCs. To this day, networked PCs in macOS are represented by beige CRT monitors displaying a BSOD. But the BSOD is supposed to be a diagnostic tool, an informational screen that technicians can use to begin homing in on the problem that caused the crash in the first place; that old Windows' BSOD error codes were often so broad and vague as to be useless doesn't make the idea a bad one. Today, version 255 of the Linux systemd project honors that original intent by adding a systemd-bsod component that generates a full-screen display of some error messages when a Linux system crashes.
The systemd-bsod component is currently listed as "experimental" and "subject to change." But the functionality is simple: any logged error message that reaches the LOG_EMERG level will be displayed full-screen to allow people to take a photo or write it down. Phoronix reports that, as with BSODs in modern Windows, the Linux version will also generate a QR code to make it easier to look up information on your phone.
The systemd-bsod component is currently listed as "experimental" and "subject to change." But the functionality is simple: any logged error message that reaches the LOG_EMERG level will be displayed full-screen to allow people to take a photo or write it down. Phoronix reports that, as with BSODs in modern Windows, the Linux version will also generate a QR code to make it easier to look up information on your phone.
For boot errors only (Score:4, Informative)
FYI, as part of systemd, it's only for boot errors.
Linux does still have the kernel panic for other errors.
Re:For boot errors only (Score:4, Insightful)
Re:For boot errors only (Score:5, Funny)
Linux does still have the kernel panic for other errors.
Like for when it realizes it's running systemd? :-)
Re: (Score:2)
Ironically if the kernal panics during boot the only way you're logging this is if systemd is running, ... or if you have an ancient serial console recorder but who the heck has one of those.
Re: (Score:3)
How does that work if your console is a serial port?
Re:For boot errors only (Score:5, Funny)
I'm sure there's an ascii driver for QR codes. BUT....Who in their right mind would trust a QR code generated 1) by systemd or 2) displayed after the computer has crashed.
Re: (Score:3)
Who trusts a QR code, period?
How do you know somebody hasn't replaced the QR with a malware link? Give me a f***ing plain text URL instead.
I hate QR codes.
Re: For boot errors only (Score:3)
The problem with plaintext is you have no way to copy-paste the URL in this situation. You can always just use you phone to decode the QR code and look at the URL before following it.
Eh, typical for Systemd (Score:2, Redundant)
Systemd is pretty much the opposite of the Linux philosophy, so it's not unexpected to get a Windows-style BSOD. Not saying it's a bad idea, and it should not affect me as I only have to do with systemd on headless servers, but still mildly amusing.
Re: (Score:2)
Yes it is (and that really doesn't thrill me), but someone had to do something about SysVinit.
Re: (Score:2)
Can’t wait (Score:3)
For the QR codes to point to a domain that someone forgets to renew.
Re: (Score:2)
Yep. Otr that some malware pusher has renewed after it expired.
"Cretin-level" design. No surprise to find utter crap like that in systemd.
Re: (Score:2)
Those are set by the distribution owner, actually. They're customizable so any errors the distribution can have the QR code point to a webpage to contain common issues and resolutions.
The whole point is to answer the question "Your system has crashed, now what?"
This can be useful for embedded devices as well if they crash they can show it on a screen which the user (who probably doesn't know or care it runs Linux) can then go visit for trou
Re: (Score:2)
"Those are set by the distribution owner, actually. They're customizable "
Which absolutely *guarantees* that they will wind up pointing to an outdated URL.
Windows is a cancer ! (Score:5, Funny)
Guru Meditation (Score:2)
At least be creative with it...
Another reason to stay away (Score:3)
This is just exceptionally bad technology. No surprise it takes its inspirations from Windows.
Re: (Score:3)
Yeah, but they're ALL stuck with it. All the major distribution choices are blocked by systemd.
Re: (Score:2)
And that is the problem. Gentoo, for example, nicely demonstrates that you do _not_ have to make your distro systemd-specific when you start to support it. Too many distro maintainers (including Debian) have apparently gotten fat and lazy and pretty stupid.
I have absolutely no problem with systemd as an option (which I will then never use). I have a massive problem with it being the only or only supported option.
Re: (Score:2)
Too many distro maintainers (including Debian) have apparently gotten fat and lazy and pretty stupid.
You can use another init system on Debian, although it is not supported by the installer. But it takes a bit of fiddling which is already done for you in Devuan, so there's no reason to install Debian if you want to not use systemd.
But yes, maintainers have gotten stupid. That is literally the main argument for systemd, it means you don't have to write init scripts in theory. In actuality it means you have to write fewer of them, but still some of them, because systemd cannot do all the things with a unit f
a reminder (Score:4)
Devuan Linux is Debian without the systemd, and with the various tweaks you need for the system to function properly without it.
I am running Devuan 5 with root on ZFS on my desktop, with X11 and KDE, and ifupdown. It works just exactly how I expect. The boot speed is absolutely fine. Errors actually make it into the logs reliably.
You do not have to give up the systems that work for the systemd that doesn't.
Re: (Score:2)
...It works just exactly how I expect.
That is exactly why systemd was such a godsend. SysV worked exactly like I expected: clumsily and unreliably. It sucked soooo bad compared to systemd, which made a Linux desktop more usable for regular people.
Re:a reminder (Score:4, Interesting)
Re: (Score:2)
"Regular people" are not dealing with systemd or SysV, both would be hidden/abstracted away from them.
Re: a reminder (Score:3)
What regular people? You mean on Slashdot? This is not a group of regular people.
Re: (Score:3)
...systemd, which made a Linux desktop more usable for regular people.
And this is why I think it may be acceptable for desktop or single-user systems but it's not necessarily ideal for servers.
If there's an issue with initialization on a server, I'd prefer to troubleshoot and put it in a state where I can see the order of operations. I don't trust that is possible with systemd.
Re: (Score:2)
And this is why I think it may be acceptable for desktop or single-user systems but it's not necessarily ideal for servers.
It caused me problems on my single-user desktop system, which is why I went to Devuan. I had boot issues that were too hard to troubleshoot for the zero return I got from systemd.
Re: (Score:2)
Indeed. While I am still mostly on no-systemd Debian, if that ever becomes too clunky, Devuan is a definite alternative. I also have a non-systemd Gentoo running, but Gentoo requires a bit more set-up than other systems. It is done cleaner though, less auto-"magic" that can very easily turn into "auto-mess".
Re: (Score:2)
What does it use for the init system? I would love a distribution that doesn't use systemd, but if it's SysVinit, I'm not interested. That's amateur hour.
When I first started using Unix, back in the 80s, I spent months going through the manuals, trying to figure out how the startup process handled dependencies. I mean... a multitasking system like Unix must have a parallel startup process, right? Right? I was pretty disappointed when I found out how it really worked
Re: (Score:3)
What does it use for the init system? I would love a distribution that doesn't use systemd, but if it's SysVinit, I'm not interested. That's amateur hour.
sysvinit by default, you can install runit or finit. Or cloud-init if you need that. I am happy with sysvinit. It's just me on my system, and it can't do two things that truly tax it at once anyway because (hilariously enough) I run out of VRAM. It costs quite a lot to get much more, though.
When I first started using Unix, back in the 80s, I spent months going through the manuals, trying to figure out how the startup process handled dependencies. I mean... a multitasking system like Unix must have a parallel startup process, right? Right? I was pretty disappointed when I found out how it really worked.
Say hello to my little friend [fsf.org]
Next for systemd (Score:3)
After they finish the BSOD screen, the systemd team's next project is to add Clippy to the boot sequence.
Re: (Score:2)
Yes.
What do you mean? (Score:3)
What do you mean, when a Linux system crashes?
Re: (Score:2)
Microsoft Contribution (Score:5, Funny)
I see Lennart Poettering has been busy since becoming employed by Microsoft
SystemD (Score:3)
Can we just purge all this shit once and for all?
I want my fucking readable plain-text scripts back that did everything and DIDN'T fuck up my machine on a regular basis ever since they were introduced.
Everything I want to change on systemd turns out to be impossible (renaming certain device nodes) or involves completely handing over the system to systemd in all aspects (booting), or having to entirely remove systemd's handling of something it has no need to be touching in the first place (e.g. DNS).
Re:SystemD (Score:4, Insightful)
Thank you. When the talk turns to systemd I get so clamped up with anger it's hard to express myself. Yes. I want my text files back, dammit. I want to choose my own tools. I want things simple. Every time I have a problem which turns out to be something else systemd hijacked, I get even angrier. And as a computer professional, it's not like you get to choose not to use RedHat at work, so it's impossible to avoid that pile of shit systemd.
And it seems the Linux people who think systemd is fine are also MCSEs or whatever they're calling that now and think Windows and Linux are both just great.
Re: (Score:2)
FreeBSD is pretty nice in this respect (not nice in other respects). You can edit text files for configure the boot up. You can also use the "sysrc" utility to modify lines of those files in a safe way or from a script. Like if you have some deployment scripts or Ansible that sets up a bunch of machines. And man rc.conf(5) lays out the general structure of all these config options, and gives you a way to load specific drivers, set interface names, configure networking options, etc.
P.S. I'm running FreeBSD +
Re: (Score:2)
So what is their reasoning behind binary logs?
Re: (Score:2)
Binary logs are better.
Journald can log a myriad fields, all clearly delimited. With binary logs there's no need to parse your logs with a regex, since every field is unambiguously extractable. There's no problem with messages containing multiple lines. You also can display time in whatever format you find convenient, and log binary data without any issues.
You can choose output formats and fields and dump in JSON for instance.
It's also a format with cursors which means it's trivial to resume parsing.
Journal
Re: (Score:3)
Binary logs are better.
... goes on to list a bunch of shit no one ever wanted in a boot log.
Dumping log data out to JSON with binary data in it? This isn't a log from Adobe Dreamweaver or the latest image generating AI; This is kernel boot information. WTF is JSON doing in this conversation?
You also can display time in whatever format you find convenient
WTF are you on about? Are you implying you couldn't reformat time strings before?
$ date -d "2023-11-15 08:23:45" "+%A %B %d, %Y %H:%M"
Wednesday November 15, 2023 08:23
Cursors so you can resume processing??? File and line number. Done. My editor keeps track of all my last positions in files (vim). You need curso
Re: (Score:2)
This is 2023. The log is for everything, both boot logs and applications, and servers. I don't see why my log system shouldn't be able to log say, HTTP Referer as its own field, and let me search by it.
No, you shouldn't need to. "j
Re: (Score:2)
It's so strange that people who don't like systemd want to banish it back to Hell and people who do like systemd are like "it's fine if you want to use a sysv init, great that you have the choice!"
Maybe there are rabid aystemd enthusiasts but I haven't seen them.
Very odd situation.
Re: (Score:2)
"Maybe there are rabid aystemd enthusiasts but I haven't seen them."
Oh? Then why was systemd shoved down the throats of the user bases of every major distro? Someone had to be rapid enough to push forward despite opposition. On the bright side, now they've introduced BSODs to go with the windows-like crashes brought by the windows-like features.
I blame the Solaris folks. Systemd reminds me of a crufty old binary Solaris system rather than the modern neat and trim everything is text linux systems which rende
Re: (Score:2)
Same reason Pulseaudio was shoved down our throats even though it was incomplete and quite buggy. For some reason, every distro maintainer creams over whatever Leenart releases even when it's an incomplete pile of crap.
Re: (Score:3)
It's so strange that people who don't like systemd want to banish it back to Hell
I don't want to banish anything, but they made it the only way to opt out of systemd. Say what you will about all the other init systems, but they could replace one another, rather than superseding the others and causing lock in. Making Gnome depend on logind which depended on systemd... what the hell guys? That's not the way we (used to) work here.
will the colour be user selectable? (Score:3)
In win3.x and win95, the BSOD colour could be user-defined via config file. Will systemd allow colour customization? Or will it be white characters on light blue background forevermore like in WinNT?
Re: (Score:2)
Only if there is a god. :D
Re: (Score:2)
LOL, U R a true old school Winderz user if U remember this!
What happens when pid 1 crashes? (Score:3)
Let's make a very complicated program with lots of moving parts and run it as our init process (pid = 1). Oh wait, when this program terminates the kernel panics, as is by design because a basic requirement in Unix and Linux is that the init program does not exit for any reason.
journalctl is great. Except when you're trying to debug something important and need to look at a man page to work out how to view the end of the log instead of the beginning. a normal person would use familiar tools like less/more/most, cat, grep, etc to poke around in /var/log. That's not how things work anymore, you have to use some over-engineered software that is perhaps very enterprise friendly but not very workstation friendly. P.S. check out -f, -n, and -g for journalctl that will get you through 99% of the problems.
There is a lot of scope creep with systemd. It expands to satisfy every new requirement that the major vendors push onto it. Most of these requirements are about getting Linux into the data center because the licensing is much more lucrative for everyone. The year of the Linux desktop has already past us by, and this OS is moving further away from us users and more towards the data center.
That's not to say that you can't use Linux on a desktop. Many of us certainly do. But the ability is disappearing for one person to slap together something that solves their custom problem, to put a band-aid over something that needs a workaround, or add their own customizations that the distro vendor didn't pre-package.
Re: (Score:2)
This is because developers wanted to get rid of those pesky admins who kept saying no to all the crazy insecure unstable pre-alpha garbage they wanted to run and deploy in their completely fresh off the press code they want to deploy to production.
They not only don't want linux on the desktop. They want every linux instance to be a completely fresh linux instance that only serves as a thin wrapper for their application and is spun up by code.
My work computer never had a BSOD (Score:2)
I'd changed it to a sort of mauve color, so it was an MSOD, and an MS OD is so much a better acronym
Okay...not sure it's that useful (Score:2)
In my admittedly limited experience of having to diagnose errors that cause a kernel panic, the relevant bits usually end up in the kernel log a few screens up from the straw that broke the kernel's back.
Makes since... (Score:2)
First systemd introduced crashes to linux, now it introduces the BSOD to go with those crashes. Imagine what fancy new features systemd will bring over from the POS windows world next!
I keep see "when are we going to get rid of..." (Score:2)
I keep see "when are we going to get rid of systemd and bring back _x_".
First, I F-ING HATE systemd.
Second, YOU can 'bring back" whatever you want at any time. Stop using it. There's non-systemd distros that you could use.
Systemd is never going to die as long as everyone uses it. Griping about it and then still using it does no good.
Re: (Score:3)
Why??
Re: (Score:2, Insightful)
I have asked this same question many times. Why do some people hate systemd so much?
I have gotten many different answers and many of them were so technically specific as to be almost indecipherable. They seem to boil down to "I don't approve of the technical approach taken."
Some that I remember and did understand were along the lines of "its one big giant monolith, instead of several separate independently-maintained components. That violates the core design philosophy on which Linux is based." And also
Re:Oh no (Score:4, Interesting)
You are asking a manipulative question. Marks you as an asshole.
Nobody "hates" systemd. You do not hate bad tech, you just do not use it. Same as you do not hate a rabid dog. You just kill it.
Who gets hate is _people_ that want to push bad tech on other people.
Re: (Score:3, Insightful)
You are asking a manipulative question. Marks you as an asshole.
Do you really think it's him who sounds like the "asshole" here?
Re: (Score:3)
Do you really think it's him who sounds like the "asshole" here?
Making excuses for shit behavior is a shit behavior. Calling out shit behavior is necessary.
Re: (Score:2)
Actually there are people who hate systemd. They were happily running some distro, until said distro forced systemd on them, so they had either to accept systemd and live with it, or change distro, lose habits, scripts, knowledge.
Re:Oh no (Score:5, Insightful)
I have asked this same question many times. Why do some people hate systemd so much?
I have gotten many different answers and many of them were so technically specific as to be almost indecipherable. They seem to boil down to "I don't approve of the technical approach taken."
Some that I remember and did understand were along the lines of "its one big giant monolith, instead of several separate independently-maintained components. That violates the core design philosophy on which Linux is based." And also "It logs in binary. That makes the logs harder to read." There are plenty others I don't remember though, because they just didn't seem to actually matter.
I was expecting to hear things like "It slows down all system processes by 20% and breaks NVIDIA compatibility, so you can't play games" or "it phones home to china and sends encrypted data there and freezes your system if you try to block this in your router" or "it requires you to use some non open source binary somewhere important" or "it eats up half your system memory" or something else that would clearly and directly be harmful to me. But I never found answers like that.
It was a solution in search of a problem. Linux (and unix for that matter) already had tools for manipulating plain text log files. People used them for decades and they worked as designed. So now you have binary logs and a new suite of tools to do the same job as the old tools. What exactly was gained? Once in a while I look through the "journal" and 99% of it is worthless garbage. It's actually too verbose and anything important is lost in the noise. The problem is major projects that should be cross platform are dependent on systemd to work.
That brings me to networking. Without googling, how would you set a static ip address on a modern Linux distro? Does it use Network Manager, Netplan, Connman, or something else? Ah you can use the "ip" command, You know the program that does the exact same thing as ifconfig but with a slightly different syntax, because fuck you we invented this and you didn't.
Re:Oh no (Score:4, Insightful)
It was a solution in search of a problem.
That's emphatically not true, as evidenced by other Unix-based OSes writing replacements for their init systems. For a pretty balanced view of how all of this came about, check this out from a few years ago: https://www.youtube.com/watch?... [youtube.com]
Once in a while I look through the "journal" and 99% of it is worthless garbage. It's actually too verbose and anything important is lost in the noise. The problem is major projects that should be cross platform are dependent on systemd to work.
This doesn't really have much to do with systemd or the journal. The journal just captures the logs, it doesn't have any say over what programs write to it.
Ah you can use the "ip" command, You know the program that does the exact same thing as ifconfig but with a slightly different syntax, because fuck you we invented this and you didn't.
This also has nothing to do with systemd - ip comes from the iproute2 package. FWIW though I do agree, as a user of ifconfig it would have been nice if it could have been adapted/enhanced instead of completely replacing it.
Re:Oh no (Score:5, Informative)
That's emphatically not true, as evidenced by other Unix-based OSes writing replacements for their init systems.
Most of the other init system replacements are not only analogous to init more than systemd (they don't throw the kitchen sink in) they are also typically compatible with init. They even read the same config files. Only a subset of them don't and most of those are dead, like upstart.
Meanwhile sysvinit plus startpar gives the principal benefit of systemd that users care about (reduced boot time) without all the other BS.
Re: (Score:2)
I'm not watching a 47 minute long video about systemd. An init system has no business interfering with my resolv.conf file.
Re: (Score:2)
That brings me to networking. Without googling, how would you set a static ip address on a modern Linux distro?
Install a desktop and use a GUI widget. The Windows way.
Re:Oh no (Score:4, Informative)
I have asked this same question many times. Why do some people hate systemd so much?
Because it's yet another configuration system, and it invariably doesn't replace all of the other configuration systems on themachine. So by the time all is said and done, I usually end up dealing with network configuration chances that are split between some RC file for dhcpcd or whatever, systemd's config files, and at least one or two other configuration systems, none of which talk to each other or share configuration.
At some point, I'd just rather have rc files that work than a systemd setup that's supposed to be automatic, but ends up requiring constant futzing to get it to do what I want it to do.
Re: (Score:2)
And my "chances", I of course meant "changes". :-D
Re: (Score:3)
Jeez. And a missing space that I didn't even see. And I can't even blame the time of day, because it was the middle of the afternoon. *sigh*
I'll blame the day. In the words of Arthur Dent, "I never could get the hang of Thursdays."
Re: Oh no (Score:3, Insightful)
Booting has never been a problem since I started to use Linux, about 20 years ago. Then came systemd and, still today, I encounter troubles when booting because of systemd being locked waiting on some stuff to happen. I had no real position about it when it went default, but now I am disatisified about it. Maybe it's my distro (Debian), but I start to think it's systemd.
Re: (Score:2)
But, but, but.... with systemd your laptop boots 10-20 seconds faster!!!
Re: (Score:3)
:D
I actually tested this!
Installed Debian 11 using sysv init onto my main laptop (an old Lenovo T420 with a 500GB spinning rust HDD) and it booted in under 6 seconds...
My PC is slightly faster, and that uses systemd and an SSD. I was like, WTF am I supposed to be seeing here?
Re: (Score:3)
I have asked this same question many times. Why do some people hate systemd so much?
And I have answered before: because it breaks with the Unix philosophy "Write programs that do one thing and do it well."
Re: (Score:2)
Yes, I included that answer in my summarized list of answers that I actually understood in my original post.
Re: Oh no (Score:2)
But this is the real, real shit.
It's not just an abstract technical point. It's a key architectural insight that helped make UNIX and its heirs strong, robust, deeply-reliable, and ultimately fixable OSes.
It _is_ UNIX at an abstract philosophical level.
Re: (Score:2)
systemd makes simple things complicated. And seems proud of it. (Well, ok, the systemd pushers are proud of unnecessary complication.)
I have seen exactly ZERO advantages to having systemd on my system. I haven't seen many problems, but I don't go digging into logs very often. When I want to, I see plenty of problems.
Re:Oh no (Score:5, Informative)
Why do some people hate systemd so much?
A variety of reasons, some technical, some social.
To begin with, systemd gave a very stark reminder how much control Redhat have over the Linux ecosystem. Systemd was pushed out early when it was still buggy as fuck, but GNOME decided to temporarily make it a hard dependency (since unpicked), so that was it, all the major distros basically had to adopt it. Even though it was buggy as fuck and broke a bunch of stuff.
In principle, it's not a bad idea. The idea of a system daemon that provides non kernel system services. In practice it's not an especially good design of such a thing.
I mean we're still apparently writing core infrastructure in C, and systemd is quite large and complicated so it's had it's share of CVEs to put it mildly. You either need OpenBSD style pedantry when it comes to code, a small and stable system with few bugs to iron out, or a language which helps you not make mistakes. Systemd has none of those, so it's got the usual litany of use-after-free and buffer overruns and so on.
It's also a very tightly coupled monolithic system. Don't listen to the, frankly, dishonest claims otherwise. Yes some parts run in separate processes, but they're part of systemd. They are not separately usable, in the same way that Linux modules don't make it a microkernel just because they don't have to be loaded. There's a debate about monolithic vs not, but the debate was frequently very dishonest. Anyway...
Speaking of C and breaking stuff and monolithicness. When it was new, the power button on my laptop stopped working. This was basically inexplicable, no one on forums or IRC (including strong proponents) was able to help. OMFG it was a fucking nightmare to debug. Eventually an update fixed it before I managed to (that was months on and off). A big monolithic system written in C is a fucking nightmare to work with especially when it's the init process. It brings the joy of kernel development to userland. The old way where there were a bunch of shell scripts you could just edit in /etc was much easier to break into and debug.
binary journals remain an annoyance point, especially in the implementation that systemd uses where they are tightly coupled to the version. It's one of those "just why" things. Generally you need logs to fix something that went wrong and this just makes it harder. All of these things can be worked around but it just adds extra friction here and there.
The tooling is unstable and a big old middle finger to shell script users. Here's an actual quote from the manual for one of the commandline tools:
Re: (Score:2)
They don't. You're getting an incorrect impression because you're in an echo chamber.
Slashdot on the whole is full of crusty old greybeards. People who somewhere around the 90s thought Unix was the best thing ever, and since have not found any new development in the industry worthwhile. They still think of computers as "pets", when the industry has long moved on, and think the Unix Philosophy is all there is to making good software.
People more in tune with the times are found elsewhere.
Re:Oh no (Score:4, Interesting)
> I don't approve of the technical approach taken.
Yes. That is the reason. Also the difference in philosophy. The so called technical approach taken by systemd developers is, well, manic and crazy while the philosophy is "old stuff is crap".
It would be fine if we could excise systemd from the system, but we can’t. Because systemd was designed crazily, it is as embedded into almost all distros like internet explorer was embedded in windows. If sysv init was to be replaced, then it’s simple, a correct way to do it is to define the specifics of an agnostic interface/API for anything to slot in and do the job. A standardised interface to the kernel and anything else concerned, then as long as systemd is developed to use that standard interface, it can do the job. But if anyone wants to remove systemd, for say S6, even on a whim because say systemd has a zero day and you need to ditch and switch immediately, well S6 can simply be installed to totally replace systemd.
Now S6 won’t actually replace everything, it is not supposed to, as it is not as bloated as systemd, which has feature creep as a feature, so other services will be needed to fill the requirements alongside S6, so you need logging too, well install rsyslog! You need session management? Install a session manager, in fact if it is done correctly, you could even use the systemd session management services, as long as everything implements the same interface then the sysadmin has the freedom to tailor the entire system to run any service that provides what is needed. Of course there may be bugs between daemons but that’s normal.
However, we don’t have that. We were told that sysvinit needed replacing and we got a lump of bloated code to do it. Code that follows a monolithic philosophy where old standard ways of doing things are ignored for new ways that look cool. We have binary logs for example, totally unneeded and next to useless not to mention bloated. The standard interface for everything is text, unstructured or not. Just text. Highly compressible, index-able, searchable and manipulate-able. Binary logs have none of that. Turn them off? Well you can’t, it doesn’t matter if *you want* to take control and have rsyslog provide logging for everything, you are not allowed to as the systemd developers say you are stupid for wanting that and binary logs are so embedded in the system that they are NOT optional.
So you have a lump of inefficient code that is bigger than the already bloated kernel code... developed by minds that have no clue what they are doing and tell you what you will like and if you don’t like it that say you are an old fart or you have the option not to use the code even though it eventually will creep more and more into being mandatory.
Replacement of sysv init could have been done correctly, leaving us with a modular system where the new idea of having competing init systems could have been a vibrant and dynamic state of affairs. Now however we have to figure out how to replace systemd in order to have that state of affairs. Not good.
We traded simple and well understood code for a bespoke lump of crazy code, complex and ill understood, with dependencies that make a bowel of spaghetti blush, written and developed by laptop users who dont know about servers existing, who care not about the unix way or the kiss principle, who think "grep" and "find" and other small tools should be merged into one uber large lump of code that eventually becomes "systemd-userland" or whatever. And to top it all off, developers who when you or other "experts" like Linus Torvalds offer criticism, response with "wont fix" or other numerous childish behaviour, probably because I think they are very likely all kids anyway.
So there you have it. Systemd has become the most recent necessary evil. Wayland is trying to purge X.Org, which although has amazing features and a great design, is also bloated and full of stuff that simply nobody really under
Re: (Score:3)
And also "It logs in binary. That makes the logs harder to read." There are plenty others I don't remember though, because they just didn't seem to actually matter.
With systemd the act of logging typical ssh background spam consumes more processor time than anything else the system does. Searching logs was previously an instantaneous operation using grep now it takes forever, sometimes minutes.
These things matter to me. It's a needless waste of both system resources and my time for no reason.
Re: (Score:3)
And then you have to hook in using special tooling and reproduce the error. Useful if you like pain, but it certainly not very helpful. It also contains a QR code that contains some useful info. I suspect the newer Windows QR doesn't too.
Anyway, this doesn't replace that at all, it just augments it and makes IT pro's jobs easier.
Re:Oh no (Score:5, Insightful)
Here’s a thought. Instead of the QR code, just display plain fucking text.
You know eventually they will tire of /etc and decide to roll it all into a flat database file.
Re: (Score:2)
Here’s a thought. Instead of the QR code, just display plain fucking text.
You know eventually they will tire of /etc and decide to roll it all into a flat database file.
Re: (Score:3)
Here’s a thought. Instead of the QR code, just display plain fucking text.
Have fun transcribing something like this without mistakes:
pci fd90:00:00.0: BAR 4: assigned [mem 0xbffe16000-0xbffe16fff 64bit]
(This is obviously not a fatal error, just an example of why you don't want non-copyable diagnostic messages.) Better and more prevalent OCR would solve this problem though, but even a popular framework like Microsoft PowerToys ships a OCR tool that can't even recognize easy text, let alone small or blurry text.
Re: (Score:3)
Re: (Score:2)
You can automatically translate a QR code into text without any transcription error by using a phone. Don't be dense.
Re: (Score:3)
Phone?
My phone doesnt have a camera. In fact no cameras can be used on site for security reasons.
Do I have to draw out the QR code onto graph paper?
Whoever thought a QR code is a smart error reporting device is as dense as the idiot who thought binary logs files that are multiple times bigger than the text equivalents was a good idea.
The design of systemd astounds me. How can anyone be that stupid?
Re: (Score:2)
The design of systemd astounds me. How can anyone be that stupid?
If you think systemd is bad, let me introduce you to the Windows Registry, particularly the Windows 95/98 era.
Re: (Score:3)
The design of systemd astounds me. How can anyone be that stupid?
Right!??! Thank you.
Every time I'm faced with the current state of systemd and/or its future plans, I feel like there must be a conspiracy going on. All the things that old school Unix/GNU/Linux folk hated about Windows seem to be their core goals. BSOD!?!! Are you fucking kidding me? It's gotta be the longest running joke there is about WIndows! FFS, how did we let this happen? Binary logs!?!?! WHAT?
Perhaps the worst part are the fanbois that come running in defense of systemd. They generally sound EXACTLY
Re:Oh no (Score:5, Interesting)
If you think systemd is bad, let me introduce you to the Windows Registry, particularly the Windows 95/98 era.
systemd will get to that eventually. binary config files is the next logical step "after" binary log files.
The binary log files wouldn't even be a problem if systemd would ALSO do text log files at the same time. But they refuse to do that so they have to be passed off to another log system, with the obvious result that when your system is panicking some of the messages may not in fact make it into the text logs. They will be in the binary logs, which you need a special program to parse. Flat text config and log files were ADVANCES of the UNIX system. Before UNIX having config and logs in more structured files was the norm! systemd represents a gigantic step backwards in computing technology. It appears to be a deliberate move to make Linux as shitty as Windows.
Re: (Score:2)
Re: Oh no (Score:2)
That assumes that a lot of shit works right after a kernel panic has already happened, which is a very bad assumption to make.
The nice thing about a QR code is you can save yourself the step of transcribing up to 3KiB worth of text. It's a hell of a lot more useful if it's available.
Re:Oh no (Score:4, Interesting)
> Here’s a thought. Instead of the QR code, just display plain fucking text.
Exactly. Whoever thought of that idea is a great explanation for all the crazy crap that systemd developers do just because its coolz.
A QR code is totally useless over a serial console for example.
I have worked at sites where a camera is a strictly controlled device, you can’t even have one on your mobile if you want to take it around with you. So how do you decode the QR code? Do I draw it onto graph paper, then leave the site to point a smartphone on it?
And then, it's a QR code. Totally human unreadable. How do I know that it is legit? That it's not a QR code generated by a malicious bit of code? Do people still really scan a QR code and just let their device process it automatically? In this day and age? I really hope that all phones surely show the user where they are going and ask for confirmation!
What if my phone is compromised by this QR code? I mean when Android can be pwned by simply receiving a bad MMS message without me even having to open it, why would anyone who cares about security scan unreadable hieroglyphics?
Text, plain text, the “universal interface”. Oh and you can’t even copy and paste text off a website these days without pasting into a text editor first to confirm you copied what you were seeing.
Re: (Score:2)
debian already went the systemd way so i wonder how they would not incorporate this new feature of systemd too.
besides, not a fan of systemd at all but i don't see what's fundamentally wrong with the idea of sending error information to the the screen ... it's definitely not the worst idea that linux has copied from windows, and i'm looking forward to all the good jokes coming in.
anyway, i liked it a lot more how tramiel did it on the atari st: https://pbs.twimg.com/media/D3... [twimg.com]
Re: (Score:2)
The main problem with sending error messages to the screen is that the screen is transient. Requiring that the screen be in graphic mode rather than text mode, now *that's* a problem. Presumably the messages are duplicated where you can look them up, but that may be optimistic.
Re: (Score:2)
The main problem with sending error messages to the screen is that the screen is transient.
as a last resort that might be of use for the user but doesn't exclude other logging/diagnostic procedures i see no harm ...
Requiring that the screen be in graphic mode rather than text mode, now *that's* a problem. Presumably the messages are duplicated where you can look them up, but that may be optimistic.
... but ofc the devil is always in the details. tbh i haven't really looked into those but imo that would be just a design/implementation issue, and every good idea may be a disaster if implemented improperly. to continue with the honesty, i lack the detailed knowledge about how modern hardware interrupts go so ... i just hope they do the right thing!
Re: (Score:2)
you don't need graphics mode to do a QR code [github.com].
I'm not even anti QR code, this is one of the least offensive things systemd has done, so long as you can disable it.
I use Devuan with sysvinit and I have graphic boot turned off so I can see the console messages. Obfuscating the boot means just one more step to go through before you can start troubleshooting. That's for wanks and noobs who care more about style and functionality.
Re: (Score:2)
Hey whoever modded this down to -1 with "Troll." Why? Accidental? Big fan of Debian and you perceived a sarcasm tag where there was none?
Wish I had mod points.
Re: (Score:2)
:D Yeah.