MX Linux Forum

anticapitalista wrote: Mon Aug 05, 2024 3:49 pm @AVLinux There are several init systems around that have been recently developed to try and move on from the weaknesses of sysVinit (which was almost 100% universal on all linux). Why do they not get any 'press coverage' let alone corporate backing?

For best practice. A hegemonic solution is never a good solution. It is an empirical rule. especially at a low level that centralizes everything.

anticapitalista wrote: Mon Aug 05, 2024 3:49 pm @AVLinux There are several init systems around that have been recently developed to try and move on from the weaknesses of sysVinit (which was almost 100% universal on all linux). Why do they not get any 'press coverage' let alone corporate backing?

Hi @anticapitalista ,

I'm getting the feeling that I may be getting on your nerves in your responses to me earlier in this thread and currently... Nothing I've said is intended to convey that systemd is superior in any way or that I'm a particular fan of it over anything else. Yes, I need it to support Enlightenment and to be fair I also have nothing against it but my post above is was after observing and getting drawn in to some more or less "tribal" types of comments I remembered I should try to be neutral about it (until better informed of it and the other init systems) because I don't personally have direct insight or knowledge as to the intentions of the developer nor do I have full understanding of it's technical details (I would assume that you do..). To clarify earlier on when I agreed with @AK-47 that "There is more than enough fragmentation (under the guise of choice)" I was agreeing that this is an undesirable reality of competitive ideas without a clear mediator, that comment had absolutely nothing to do with inits directly, it was simply agreement of his assertion that we indeed often have fragmentation under the guise of choice, I don't know how anybody who has used Linux for more than a few weeks hasn't bumped into that reality, it's baked into the cake and yet most of us love it anyway..

As far as getting coverage, it's a rhetorical question, yes many things of true merit are suppressed by those in power, I'm looking southward (as an observer) at an election with 3 candidates yet one of them is completely not covered by the mainstream media sources so I fully get what you're saying. I'm also certainly not saying the promotion of systemd is all for altruistic reasons, I don't know, and if I don't know I shouldn't pile up on one team or the other and I feel casting uninformed judgements about what we are actually talking about isn't doing the conversation any good whether it's myself or anyone else..

AVLinux wrote: Mon Aug 05, 2024 2:53 pm At the end of the day Poettering and these people are making decisions based on how best to achieve the personal goals they have and to best solve the current bug or problem that they are tasked with untangling at any given moment and hopefully preventing from happening again. Perhaps these moment by moment decisions unfold in a better direction and perhaps they simply unfold in yet another way of doing the same thing, in Linux we don't have a top-level Project Manager to separate the wheat from the chaff so to speak so when you use Linux you are kind of signing up for the benefits and deficits of anarchy simply by the unrestrained nature of it's development. There doesn't have to be ill intent involved at all for this to be occasionally frustrating but at least I try now to take a breath and give the designer/developer the benefit of the doubt for simply problem solving the best they know how with skills that I'm not qualified to question or judge..

The issue/concerns with Poettering and SystemD have been borne out by the actions and directions taken since then. SystemD was meant to solve the following three problems
1) Dependencies during services/daemon startup
2) Parallel start of services leading to faster boot time
3) Tracking of services crashes and restarting them.
The issue was that the approach taken was very similar to what Windows does and did not adhere to the Unix/Linux philosophy of excelling in doing one thing and one thing alone. In Software architecture there is a concept of separation of concern. In simple terms it means do the thing that you are meant to do and do not step on someone else toes. Since then SystemD has morphed into something else entirely
1) Jornaling of events, aka similar to Windows Event Manager
2) User Login and Session login
3) /usr merge where the solution was said that certain Unix OS have done it so should be done by Linux too and dubious benefits were laid out for its need
4) Creating unnecessary dependencies where none needed to be exist. GNOME, udev, etc. The list keeps on growing
5) Creating an alternative of SUDO
6) etc, etc. etc. The list goes on and on

What Poettering and his supporters have managed to achieve is a level of unparalleled homogeneity in the Linux ecosystem which had thrived on diversity. And that too for little benefit. Ask any genetics expert he will tell you the benefits of diversity and perils of lack of genetic diversity among a species. Ask any political science student and they will tell you the healthy need for variety in approach and implementation. Ask any biologist which forest is more strong and resilient, one which has a single type of vegetation or which has numerous different types and species of vegetation.

The issue with SystemD is that in its guise it is converting everything in Linux similar to a single distro, RedHat and bringing it way close to Windows. I am against things being forced down just because it has to be done a certain way. And especially when the decision to go with SystemD was taken in a vote which till date is contested.

Disclaimer: I have nothing against SystemD. Or using it. Or against Windows. Nor do I have them being used. Nor do I have against the corporate backing of SystemD.

AVLinux wrote: Mon Aug 05, 2024 2:53 pm in Linux we don't have a top-level Project Manager to separate the wheat from the chaff so to speak so when you use Linux you are kind of signing up for the benefits and deficits of anarchy simply by the unrestrained nature of it's development.

That's not true. We have a top-level project manager in Linux and his name is Linus. It's his project. It's his kernel. If Linus decides your code doesn't belong in his kernel, it doesn't go in. If he decides he doesn't like your code being in his kernel, it comes out. systemd, largely, does not modify the Linux kernel purely because Linus won't let it. This means that systemd is relegated to living above the kernel level, where it is actively consuming everything it can. You're free to build Linux without a dependency on systemd. The kernel source code is free. Even the major distros like Fedora and Debian offer their codebases for free so you can remove systemd from it as some distros have done, or fork the code from a point in the project's history before systemd began to metastasize. Linux is not, precisely, an anarchy as there is no clear indication of self-rule. Power in the Linux ecosystem exists among Linus, who acts as a largely aloof philosopher king, and a Landsraad of fiefdoms who act as the great houses of muscle in the realm making the real decisions that affect their users every day. There's no way for Linus to kill off a major project like, oh, say, CentOS, so the other major corporations more or less prey on each other like junior high school bullies looking for attacks of opportunity as well as popularity and acceptance. The tech world these days is less about killing companies like it was in the 1980s and 1990s and more about killing off specific ideas. Remember Mir? Remember Upstart?

LinuxSpring1 wrote: Tue Aug 06, 2024 1:22 am The issue/concerns with Poettering and SystemD have been borne out by the actions and directions taken since then. SystemD was meant to solve the following three problems

systemd was borne out of one problem to solve: Poettering's dissatisfaction with init. As is his standard modus operandi, he disliked the previous software's design, half-baked his own solution, and then turned it loose upon the world and called it good. He's repeatedly done this with zeroconf (Avahi) and audio (PulseAudio), to name a few. He is a programmer of moderate skill who is young and brash and gravely misjudged the complexity of system service management and is now trying to code his way out of an intractable problem. There is no logical path you can take to start with "startup should be faster" and end up at "I re-wrote sudo" with a detour through "I also wrote a DNS resolver from scratch and my own version of cp". He is clearly in over his head and chose a long time ago to just keep going. Gotta admire his chutzpah.

LinuxSpring1 wrote: Tue Aug 06, 2024 1:22 am The issue with SystemD is that in its guise it is converting everything in Linux similar to a single distro

Poettering pivoted long ago to calling systemd his attempt at providing Linux standardization. And wanting consistency among distros is a laudable goal, but only if the code that provides it is sane, safe, and doesn't have 2.2 thousand open issues on GitHub and produce about a dozen full-blown CVEs a year. Poettering built a new DNS resolver. It's buggy and insecure. He decided on a new binary logging format. It's buggy and insecure. He built a user management infrastructure that gives UIDs root privileges if the username starts with a digit. He wrote a replacement for sudo. It's buggy. And insecure.

What do you gain from systemd? A slightly faster startup (sometimes), your machine will hang trying to shut something down (sometimes), you can't grep or tail -F anything in /var/log anymore, time synchronization problems, and you can start services with a bespoke unit file if you know how to format one. This last one is oddly similar to how daemontools works, only daemontools just needs a shell script. A lot of the daemontools approach was reimplemented, poorly, in systemd. A better replacement would be something like s6, which can also run as PID 1 if you really, really want it to be. The advantage of s6 over systemd is that it's (a) far less buggy and (b) if you use it you never have to deal with Lennart Poettering.

systemd isn't just "converting everything in Linux". It's actively making Linux a less stable, less secure platform and as an unapologetically crass power grab by a spoiled brat who thinks that calling his releases "Now with 42% less UNIX philosophy!" is a huge slam dunk.

DukeComposed wrote: Tue Aug 06, 2024 3:30 am

AVLinux wrote: Mon Aug 05, 2024 2:53 pm in Linux we don't have a top-level Project Manager to separate the wheat from the chaff so to speak so when you use Linux you are kind of signing up for the benefits and deficits of anarchy simply by the unrestrained nature of it's development.

...
Power in the Linux ecosystem exists among Linus, who acts as a largely aloof philosopher king, and a Landsraad of fiefdoms who act as the great houses of muscle in the realm making the real decisions that affect their users every day.

Landsraad ??? Are you a fan of Dune or Dune messiah or Children of Dune? So does that make Linus equivalent to Shaddam Corrino IV?

About the comment on S6. From what I have read S6 had two processes. PID1 and PID2. PID 1 was responsible for starting PID 2. Thats all. Nothing more. Most of the grunt work was done by PID 2. And that is what was questionable. Why have this hierarchy of processes? If unnecessary complicates things. Typically in Unix/Linux world the purpose of PID 1 is to start the system and services. Thats it. Nothing more. Once that is done, it simply goes to sleep or gets out of the way.

LinuxSpring1 wrote: Tue Aug 06, 2024 3:51 am From what I have read S6 had two processes. PID1 and PID2. PID 1 was responsible for starting PID 2. Thats all. Nothing more.

What you're thinking of is s6-supervise, which is what does the monitoring of a process to make sure that it starts if it can and restarts it if it stops without permission. s6 is a complete process supervision framework. It maintains a service tree, starts and stops those services, monitors them, restarts them if they die, handles signals, and writes, rotates and filters log files. It does not insist on replacing /etc/resolv.conf or sudo or mount.

LinuxSpring1 wrote: Tue Aug 06, 2024 3:51 am
Landsraad ??? Are you a fan of Dune or Dune messiah or Children of Dune? So does that make Linus equivalent to Shaddam Corrino IV?

Ah.. so fun to see the Dutch language put to use

anticapitalista wrote: Mon Aug 05, 2024 3:49 pm @AVLinux There are several init systems around that have been recently developed to try and move on from the weaknesses of sysVinit (which was almost 100% universal on all linux). Why do they not get any 'press coverage' let alone corporate backing?

Unfortunately people who develop good solutions to real problems don't always get this backing because they fail to seek out and convince people to be invested in it. This isn't because these engineers are flawed in any way, but those who scream the loudest are the most likely to be noticed, even if they are mad. It is for similar reasons, DOS and early Windows got by despite superior operating systems (such as OS/2 or MacOS) existing at the time. Poor OS/2 barely made a dent in Windows' market share. Same with Intel and their take on ECC RAM, which Linus Torvalds even had a go at them for.

Eadwine Rose wrote: Tue Aug 06, 2024 5:34 am

LinuxSpring1 wrote: Tue Aug 06, 2024 3:51 am
Landsraad ??? Are you a fan of Dune or Dune messiah or Children of Dune? So does that make Linus equivalent to Shaddam Corrino IV?

Ah.. so fun to see the Dutch language put to use

Dutch language? wdym?

Land(s)raad is a Dutch word.