Slashdot Mirror
In Which Linus Torvalds Makes An 'Init' Joke (lkml.org)
Long-time Slashdot reader jawtheshark writes:
In a recent Linux Kernel Mailing List post, Linux Torvalds finishes his mail with a little poke towards a certain init system. It is a very faint criticism, compared to his usual style. While Linus has no direct influence on the "choices" of distro maintainers, his opinion is usually valued.
In a discussion about how to set rlimit default values for setuid execs, Linus concluded his email by writing, "And yes, a large part of this may be that I no longer feel like I can trust "init" to do the sane thing. You all presumably know why."
In a discussion about how to set rlimit default values for setuid execs, Linus concluded his email by writing, "And yes, a large part of this may be that I no longer feel like I can trust "init" to do the sane thing. You all presumably know why."
Make no mistake, this is a turf war.
Who's in charge? The user? The kernel? Ring-0?
The answer to this is different depending on the topic. The topic here is init and who gets to say what the rlimits are and how. There are lots of other topics - random numbers, filesystems, network attach-detatch, routing etc. For all these things and many more there has been a turf war along the lines of "We will fix this in the kernel!", "Oh no you won't, we will fix this with our daemon", "Oh no you won't, my userland administration tool will fix this".
This is generally fine, but for each there will be a slashdot thread with many jerks represented.
I should use this sig to advertise my book ISBN-13 : 978-1501515132.
Don't forget the recent severity 9.8 CVE regarding invalid username handling that Poettering closed as NOTABUG. It's a trainwreck of bad design driven by an egotistic idiot.
> doing it half-assed.
Like logging. Logging is critical for both troubleshooting and security. With sys V init scripts, even if the error wasn't logged to syslog, you'd at least see it on the console instead of so often seeing nothing with systemd.
you are one of those special idiots my mother warned me about... EWONTFIX/Closed is NOT fixing...
Updating manuals to (now) state that systemd only accepts usernames adhering to: [a-z_][a-z0-9_-]*$? is not a fix.
Systemd hasn't fixed teh issue, they man paged what it doesn't like. someone creating a username starting with a 0 will still get executed as root. Even worse!!! a username with a "." in it will also do it... Periods have been permitted for ages (just not starting...) and this means if a linux machine is part of an AD it could cause issues...
https://lists.freedesktop.org/archives/systemd-devel/2017-July/039237.html
> 1. We do not permit empty usernames
> 2. We don't permit the first character to be numeric
> (This also filters out fully numeric user names)
> 3. We do not permit dots in usernames, neither at the beginning nor in
> the middle.
> 4. We do not permit "-" at the beginning of usernames (something which
> POSIX explicitly suggests, btw)
> 5. We require that the user name fits in the utmp user name field, so
> that we can always log properly about it.
Mod parent up...
Classical init was made to handle monitoring of services, making sure they get restarted if they fail but not over and over if they keep failing. This was done with inittab.
Unfortunately inittab because too limiting, especially when it came to starting order and dependencies, and so everyone abandoned it, replacing it with a bunch of shell scripts, different depending on distribution and Unix variant. Alas, the process monitoring was lost in that change, so everyone had to run stuff like monit and write a bunch more scripts.
SystemD brings proper daemon monitoring back, on steroids. It does away with stupid PID files and it handles dependencies very very well. It is an enormous leap forwards.
Alas, it also decided to solve a bunch of non-problems like logs and DNS resolution and file system mounting. Problems that already had really well tested solutions that could be relied on to never break.
(Yes, snatching STDERR from a daemon is genius. Definitely. But what was wrong with then handing the output to the syslog daemon?)
Finally! A year of moderation! Ready for 2019?
Indeed; its author campaigned to get it into every major DE and distribution, starting with a proposal to have GNOME depend on logind. [1] He later congratulated trolls who pushed it on debian-devel [2], publicly, on Google+. [3] Later, he moved to force udev to depend on kdbus (and thus systemd; though that was thwarted by LKML), adding "Gentoo, this is your wake-up call." [4] Now, what type of developer treats his fellow FOSS members like that? I'll tell you: someone looking for a fight.
No legitimate libre software project needs to be pushed to gain adoption. Perhaps you should read more about Lennart's original goals, such as "gently pushing everyone to use the same standard base". [ https://lists.fedoraproject.org/pipermail/devel/2011-June/152672.html ] This attempt to "standardize" was little more than building a solution and *then* finding problems to fix with it. In short, it was politically-motivated. It's a big part of why, predictably, it has become a Gordian knot of "modular" components. Moving away from systemd 10, 15, 20 years from now will be a nightmare; I bet they haven't thought about that, though! By keeping components 100% divorced from each other, a proper *nix system can swap components out in just about any order or combination. Those who've invested too much into systemd will need to find a replacement for each systemd "module" they actively use if they're going to get away from it. It's the classic all-in-one software architecture that looks easy to adopt, but proves extremely painful to move away from. Do you think that was by accident? If so, I've got swamp land in Arizona for sale.
If it was as you say, there wouldn't have been a big upset over its adoption. If what you said was true, literally every distro would use it and it would have little to no flaws. A big part of what people dislike about systemd is its culture, both users and developers. Before you fire out that "only technical discussion plz" bullshit, take a look at other projects. Would you want to be part of any of them if they had a shit culture? If they arbitrarily closed bugs, assuming their software is perfect? Social merit is just as relevant as technical merit; without both, you won't get users and your community will die, which will lead to bitrot and eventually, software death. Consider that the reason most people use systemd is a) distros were "gently pushed" to use it and b) most users don't pay much attention to init. With those two facts in mind, how can anyone say systemd earned that position? It was marketed and pushed, plain as day. Your willful ignorance does not contradict it.
Now that systemd architectural flaws are showing their faces (as others have predicted), the rabble are scrambling to defend the poor technical choices made by the systemd team. The biggest supporters of systemd (Arch Linux devs in particular) lied -- yes, lied -- to their users about their plans for systemd. [5] One such developer was rewarded for their duplicity with a spot on the systemd dev team: Mr. Tom Gunderson, who I've had the displeasure of communicating with personally. Even one of his fellow devs was pissed he was spreading false information: [6]
At what point will you and others realize that the systemd team is socially manipulative and does not appreciate others finding holes in their software? How many other teams would dismiss as NOTABUG or WONTFIX, like the recent '0day' username bug? They are not good libre software citizens and are hardly worth the bits their code is stored on. The Bazaar has little use for an Ivory Tower.
Perhaps you should read up on the history yourself before proclaiming others are full of shit.
[1]: https://mail.gnome.org/archives/desktop-devel-list/2011-May/msg00427.html
[2]: https://lists.debian.org/debian-devel/2012/11/msg00350.html
[3]: https://plus.google.com/+LennartPoetteringTheOneAndOnly/posts/jcCjMct3SJ3 (See Lennart's Dec 19th commen
If you're going to bitch about bugs, there are so many to bitch from. But the ones you linked seem perfectly reasonable won'tfixs.
- One single person reporting that DHCP leases don't renew even though his logs show that the client attempted renewal. If this were worth looking at there'd be thousands of confirmations. Why should the bug be fixed if it can't be confirmed?
- Defaulting to google's NTP service is perfectly reasonable given the complete lack of alternatives. As shown in the bug tracker you are specifically requested NOT to default to pool.ntp.org unless you're the vendor, and then the configuration becomes vendor specific. i.e. if your default install is hitting Google.com then maybe you should be complaining to Ubuntu or Debian or Arch or whoever decided to blindly include systemd-timesyncd without creating a proper config default for their distribution.
-NFS mounts are not a problem providing you RTFM.
-A start job is running is a perfectly sane response to waiting for a critical part of a boot process, and it has the perfectly sane action of eventually timing out. If this occurred without change then chances are it's a hardware failure. If it occured due to an upgrade then you distribution maintainer did a shit job at adding the new package. e.g. the "bug" introduced in systemd 230 which curiously only affected Arch linux.
Every software has bugs, but systemd bugs are closed EWONTFIX because the principal developer has zero clue about modern operating systems.
Actually I find the problem more with the peanut gallery who think that ever turd smeared on a bug tracker is critical or even real. Like the guys who keep quoting the "open bugs" graph of systemd without realising that some 2/3rds of the bugs being posted are RFEs.
With a server market share of more than 50% (look up Netcraft monthly stats), and a desktop market share of 1% -- so guess where the priorities are
With the servers, where management of massive logs and monitoring of running processes was a key design goal and one of the primary reasons why the likes of Debian and RHEL adopted systemd.
They want gnome and systemd is the price.
The only sad part is, if everyone here used all the energy they have to hate on SystemD, to actually fix those bugs, we wouldn't be having these discussions about how buggy it is.
Anyone who has tried has had issues waved away as someone else's problem. This also does not resolve the maintainer or responsibility, a notion which is just downright hilarious.