Lennart Poettering goes to Microsoft

mefizto said:
since the thread has already deviated from the original topic and touched inter alia on systemd, could someone please explain in simple terms the problem with systemd?
Click to expand...
The biggest reason that everyone has a problem with systemd it massively bloated monolithic project. As it has been mentioned several times, is the principal of do one thing and do it well. Systemd was originally said to be a init system; now it can't be grouped in any category. To be specific, when I am meaning monolithic; I am meaning it is all parts of it is hard coded to use each other. Anymore, the list of all the subsystems that systemd provides is an ever growing list to include anything and including the kitchen sink.

Some parts of systemd I admit that I do like and/or don't mind using; others I do not like. The biggest issue for me, is that I HAVE TO use the entire blob; you are given no choice.

As far as Poettering, I'd would bet he is working initially on the WSL2 subsystem. The reasoning is more of fixing/advancing the systemd support/integration. I know it has been desired to run more linux GUI apps, but it's been missing the systemd support that everything is dependent on and progress the wayland side to more efficiently display the GUI's.
 
Hi Jose,

thank you for the articles. If I may restate them as I understand them, there were some poor high level design decisions, cf., the first document, the discussion re the PID and the need to run some other processes at elevated privileges.

In all honesty, i do not quite understand the technical aspects of the second one, but, it seems that some features were implemented that had been billed as improvement, which either were not, cf. the discussion of "socket activation", or they were paid for by high complexity and maintenance requirements, cf. A3 and the specific heading. Furthermore, it seems that for some reason that is not quite understood from the article, it "swallowed" so to speak additional processes that might remain independent, cf. the cgroup discussion.

Hi ct85711,

ct85711 said:
To be specific, when I am meaning monolithic; I am meaning it is all parts of it is hard coded to use each other. Anymore, the list of all the subsystems that systemd provides is an ever growing list to include anything and including the kitchen sink.
Click to expand...
I am not knowledgeable to dispute your definition of "monolithic", although I was going to use the term in regards to the "swallowing" and the discussion about setting a policy how to do things, and not providing tools to do things.

Thank you both.

Kindest regards,

M
 
mefizto said:
since the thread has already deviated from the original topic and touched inter alia on systemd, could someone please explain in simple terms the problem with systemd?
Click to expand...
It's a crucial test for the Linux community, because it's the opposite of what Linux originally standed for:
systemd violates Unix Philosophy because it's large and monolithic.
And it's a scribbled piece of software full of bugs.

But as from FreeBSD point of view I would not so much be concerned about what's going on in the Linux world, especially not since Github already belongs to MS.
(Ever asked yourself what's worth paying 7.8 Bio US$ for?
To scan for useful code and talented programmers?
Anybody can do that for free. No need to buy Github for that.)

For feeling part of the FreeBSD community for me such things are incentives to more clearly seperate from Linux,
underline the definition of being something own,
and keep an eye on ensuring freedom and independency.
 
mefizto said:
since the thread has already deviated from the original topic and touched inter alia on systemd, could someone please explain in simple terms the problem with systemd?
Click to expand...

systemd is really just a poor implementation of launchd. instead of just coherently merging service related daemons; it went way overboard with its feature creep and swallowed most of what runs beside the GNU userland. Binary logs emulating plists, dbus integration emulating Mach IPC, etc. Linux NIH syndrome at its best. I think the most idiotic thing done was making systemd a dependency for gnome-session. That's really how it became a de facto standard.
 
We had discussions about that-what-shall-stay-outside at other threads here in great depth. Please, not again. It seems this train is approaching the closing station and will be there in a day or two, if not derailed on its way.
 
mefizto said:
If I may restate them as I understand them, there were some poor high level design decisions, cf., the first document, the discussion re the PID and the need to run some other processes at elevated privileges.
Click to expand...
But that was only the beginning. There were plenty of implementation bugs features like the hilarious one where invalid usernames were assumed to be root. This turns out to be user error, of course. Just listen to Mr. Poettering:
github.com

systemd can't handle the process previlege that belongs to user name startswith number, such as 0day · Issue #6237 · systemd/systemd

Submission type Bug report systemd version the issue has been seen with systemd 232 Used distribution Linux ubuntu 4.10.0-19-generic #21-Ubuntu SMP Thu Apr 6 17:04:57 UTC 2017 x86_64 x86_64 x86_64 ...
github.com github.com
Giving root to an invalid username is a feature. You're just not smart enough to understand why it has to be so.

This was only the beginning of Mr. Poettering's crusade to explain security to us.
https://twitter.com/i/web/status/1149666841494638593
View: https://twitter.com/pid_eins/status/1149666841494638593

Systemd wins top gong for 'lamest vendor' in Pwnie security awards

Epic fails and l33t pops celebrated by hackers
www.theregister.com www.theregister.com

Poettering is a talented hacker, but he's a far more talented politician. It's frankly disgusting how his sycophantic followers crawl out of the woodwork to justify the great man's latest ridiculousness:
ma.ttias.be

Giving perspective on systemd’s “usernames that start with digit get root privileges”-bug

Fire in the hole! There’s a new systemd bug that gets the haters aroused!
ma.ttias.be

Listen, if you're going to do something about invalid input, assuming root privileges should be pretty far down the list. Even crashing is more reasonable.

mefizto said:
In all honesty, i do not quite understand the technical aspects of the second one...
Click to expand...
Well, you did ask "Is it poor architecture, bad implementation?" It's hard to answer such without technical detail.

mefizto said:
Furthermore, it seems that for some reason that is not quite understood from the article, it "swallowed" so to speak additional processes that might remain independent, cf. the cgroup discussion.
Click to expand...
Oh no reason has been given other than it's "better" to have the functionality in systemd. Or at least I'm not aware of one. There's always some mumbling about systemd being "modern" and not "outdated" or some such, but precious few specifics.
 
Beastie7 said:
I think the most idiotic thing done was making systemd a dependency for gnome-session. That's really how it became a de facto standard.
Click to expand...
So what does this mean for gnome on FreeBSD? Clearly systemd isn't about to be any 'standard' on FreeBSD, so how will gnome fare here?

I've long used KDE on the {desk,lap}top so it doesn't affect me, but I wonder if some reverse pressure from *BSD gnome users might help liberate gnome from this kind of lock-in? Or are there too few to be of influence?
 
Beastie7 said:
agreed.. but it's such an easy target. 😁
Click to expand...
My inner klingon objects. There is no honor in hitting the defenseless. But yes, they are an easy target.

Jose said:
Oh no reason has been given other than it's "better" to have the functionality in systemd. Or at least I'm not aware of one. There's always some mumbling about systemd being "modern" and not "outdated" or some such, but precious few specifics.
Click to expand...
There are two types of idiots. One type says "this is old and therefore bad", the other type says "this is new and therefore better". Beware of individuals who fall under both types.
 
smithi said:
So what does this mean for gnome on FreeBSD? Clearly systemd isn't about to be any 'standard' on FreeBSD, so how will gnome fare here?
Click to expand...
In the same way gnome can works on systems that doesn't rely on systemd (funtoo for example). Some people are willing to work on it and patch to make it happens.
 
ct85711 said:
The other part that contributed in pushing Ubuntu/Debian to switching to systemd, is that Gnome made systemd a hard requirement. In the end on Linux, KDE and Gnome are the big players that overwhelming is the main DE that most people use. In the end, Debian and other distros had little choice between switching to systemd(the least trouble), alienating half the users, or carrying patches that upstream won't accept and constantly break. Even Gentoo was constantly having trouble with this, and has eventually gave up.
Click to expand...
Glad you brought this up - and it's part of the same strategy. gnome is heavily funded by Red Hat (IBM), canonical, SUSE, etc - and is a project that is overseen and developed from a proprietary perspective, by developers who work hard to restrict what the end user can do, who actively block customisations and only care about "brand presence". gnome developers are not interested in fixing bugs from systems where systemd isn't installed.

gnome project even developed their own "registry" (dconf). The two founders of gnome and developers of mono (.NET / C#) both work/worked for MS. All old news.

For me at least, Poettering trading IBM for MS is not such a big deal. Any "stink" over this coming from certain quarters is going to be because of the infamy of systemd and its lead developer - whereas Linux fans don't bat an eyelid at the many lines of code in the Linux kernel written by Intel, google, Huawei, Samsung, ARM, MS, Meta, AMD, Microsoft, Red Hat, IBM, Oracle, etc developers.

Remember: Microsoft <3 Linux ...
 
what say about systemd that have not say yet? and not be off-topic...well,this notice is not surprise for no one
systemd->closed->all power to one->$
microsoft-> all above 🤣
lucky for all...there is FreeBSD!!
 
ct85711 said:
The other part that contributed in pushing Ubuntu/Debian to switching to systemd, is that Gnome made systemd a hard requirement. In the end on Linux, KDE and Gnome are the big players that overwhelming is the main DE that most people use. In the end, Debian and other distros had little choice between switching to systemd(the least trouble), alienating half the users, or carrying patches that upstream won't accept and constantly break. Even Gentoo was constantly having trouble with this, and has eventually gave up.
Click to expand...
Yeah, they did with GNOME3. Then again I am still on the search for that person who really wants to use that pile of trash...
 
I tried to give Gnome3 a honest attempt on using it, but I just could not accept their philosophy of needing a plugin for everything. To me, it felt like it is too dependent on plugins to have the same feel and/or functionality.
 
kpedersen said:
Linux has proven that even with all of these compromises to the OS, it can still only manage <1% market share.
Click to expand...
Well, that's an interesting question. On the desktop, Linus' market share is actually a little bigger than that, about 2-3%. Meaning it's irrelevantly small. So your argument is kind of correct: Linux' design choices have not made it conquer the desktop market.

On the other hand, among servers Linux' market share is way above 90%, probably around 99%, in certain markets (like supercomputers) it is 100%. And most of those today run on a systemd installation. Microsoft alone probably has more Linux servers in it's data centers (perhaps a factor of 10 more) than there are total Linux DE machines on the planet. Add to that the other FAANG+friends, and you see why Lennart matters.
 
ct85711 said:
The biggest reason that everyone has a problem with systemd it massively bloated monolithic project.
Click to expand...
systemd reminds me uncomfortably of vax/vms, another massively bloated monolithic project. fortunately for the future of humanity, monoliths have bell shaped adoption curves, and are hard to displace until they're easy to displace.
 
kpedersen said:
Linux has proven that even with all of these compromises to the OS, it can still only manage <1% market share.

Money and marketing are the only things that can build consumer userbases; not technical merit.
Click to expand...
Well, one problem is the fragmentation throughout distributions. Then the missing binary driver interface in the kernel is another major obstacle. The other is that Microsoft Office does not run natively on it, and probably never will.

Let's not forget when Steve Jobs returned in 1998 to Apple, who were at that time short before bankruptcy, he managed to gain a very unexpected deal back then by that time: Microsoft bought a lot of Apple shares with no voting power. And Microsoft agreed to migrate Office to Mac, probably because they were back then in the antitrust lawsuit and therefore wanted to show otherwise.

Jobs on the other hand was smart enough to realise that Apple will only gain a foothold in offices if it has Microsoft Office.
 
hardworkingnewbie said:
For me GNOME3 was always nice... for a tablet. Too bad it was a desktop environment to be used on a PC. So wrong target audience.
Click to expand...
It's a dilemma: Gnome3 works fine for a tablet as long you have a tablet with a beefy processor and a good amount of ram. 😂
 
ralphbsz said:
On the other hand, among servers Linux' market share is way above 90%, probably around 99%, in certain markets (like supercomputers) it is 100%.
Click to expand...
It kind of gained this market share well before systemd and other compromises were made though. It gained it by being a "cheap and cheerful POSIX(ish) *nix(ish) clone".

The question is, does it still achieve this or is it just sheer inertia (and lack of competition) that is keeping it at the top. I would suggest Docker might be a slight hint of a symptom that people are not really liking the management of Linux.
 
kpedersen said:
It kind of gained this market share well before systemd and other compromises were made though. It gained it by being a "cheap and cheerful POSIX(ish) *nix(ish) clone".

The question is, does it still achieve this or is it just sheer momentum (and lack of competition) that is keeping it at the top. I would suggest Docker might be a slight hint of a symptom that people are not really liking the management of Linux.
Click to expand...
The most popular Linux used for Docker is Alpine, which uses Openrc.
 
You must log in or register to reply here.
Back
Top