Linux Is Getting New Network Libraries From Veteran systemd/BUS1 Developers

[cybertraveler]

Wow what is this troll spamming? 3 comments in a row, full of pure hate?

...

pal666 is always behaving like that. I put him on my ignore list so I don't get notifications about his posts.

[pal666]

Wow what is this troll spamming? 3 comments in a row, full of pure hate?

So either you are also no developer of systemd what I strongly assume, or I begin to understand how the systemd shitstorm started, I thought only the systemd haters are toxic, you are the first systemd fanboy that is so full of hate that I met.

"you do not write code" I do not write systemd code yes, but of course I am a developer. Where did I whine, stating a very reasonable opinion is now the same as whining.

i stated very reasonable opinion, you whined and posted several comments in a row, full of pure hate. i didn't tell you how you should develop your software for free, you don't get to tell other devs how they should develop software for free

[aht0]

i stated very reasonable opinion, you whined and posted several comments in a row, full of pure hate. i didn't tell you how you should develop your software for free, you don't get to tell other devs how they should develop software for free

So much bullshit. your "reasonable opinion" tends to be either personal attacks, insults or just pure demagogy.

Quoting yourself "Nobody really cares about your opinions."

[pal666]

So much bullshit. your "reasonable opinion" tends to be either personal attacks, insults or just pure demagogy.

Quoting yourself "Nobody really cares about your opinions."

lol, did you try to apply your post to itself?

[jacob]

Unfortunately for you, that will never be achieved. There will always be GNU/Linux systems which do not adopt systemd. It is simply not suitable for all uses cases (for instance, see the comments by the Knoppix lead dev) and there are lots of people who have preferences for non-systemd init systems (whether that be runit, Upstart, sysv init style or other). There is also a wide selection of software that provides requivalent functionality to much of what other systemd daemons do. Anyone who has used Gentoo in the past should be aware of this.

There will always be distros that use all sorts of stuff, but Linux really needs standardisation. It is essential that application developers can assume certain functionalities, and more importantly, can interact with the system through well-defined APIs to set/query network settings, enable/disable services, set timer tasks etc. In other words, what Linux needs is interfaces for software developers much more than "choice" for sysadmins. If someone wants to build a distro without systemd (or without the kernel, or without libc etc...), whatever, but they shoudn't expect everyone to bend backwards to accommodate them. It had worked like that in the past under the banner of the so-called "Unix philosophy" and the result has always been a total disaster. Time to put it out of its misery.

[jacob]

I am not sure what you exactly mean about this, you hope that there will no Distro that don't use Systemd? Or you want it "integrated with Systemd" despite that distributions?

I think it should not be considered a permanent part. I am much against the Systemd hate, and I think it's to some extend tribalism and to some extend also a conspiracy theory.

But there are technical reasons for competing service managers, take shepherd its written in Lisp / Scheme, first that is the better programming language but not only because I like S-expressions more, but also for other reasons, as far as I know systemd is written in C so you not only have very slow development process because between every iteration you need to compile it, you also have security problems with C, there are some security problems that other compilers / interpreters make impossible.

Also it's not Linux specific so it could be used by other operation systems like bsd or apple or heck even windows if needed, so you can attract more developers.

C projects need always extremely much extremely competent developer community behind it so that it does not end in a total disaster, look at the Hearthbleed shit that happend with this C Programm because nobody could maintain this C Mess.

But maybe I am wrong and Rust is really the new next big thing (even I don't see how something without s-tree could be any relevant ) and a new service manager in rust will become the next big thing.

There is some beauty in having options if you want 1 fits all windows or macosx is better in doing so. We should not copy them. So if debian uses Systemd I applaud them, and will defend them from haters, but if people try to put walls into the linux ecosystem that makes it much harder or impossible to use different alternative parts I am against that.

I hope they will not sacrifice functionality and ease of use (for app developers and end-users, not sysadmins) in order to avoid dependencies. Whatever it will be, it's a suite of network configuration tools so by definition it *will* be Linux specific - and besides, I really think it's time to consider Linux as Linux, not like "one of the Unixes", so Linux-specific software doesn't worry me a bit.

As for Scheme-based service managers etc., I'm not fond of the idea. A service manager is one of the most privileges processes in the entire system, and it should be as static and declarative as possible. Having instead a full-blown programming language interpreter processing config in the form of executable code, thus prone to bugs, security issues etc., all the while requiring a garbage collector, is really BAD(tm) in my opinion. The fact that old-style init was effectively like that doesn't mean it was ever a good thing.

[cybertraveler]

I really think it's time to consider Linux as Linux, not like "one of the Unixes", so Linux-specific software doesn't worry me a bit.

Linux is a kernel.

GNU/Linux is an operating system which implements a Unix-like environment as is somewhat compatible with many Unix-like standards (including POSIX).

Mac OS X also happens to implement a Unix-like environment and last I checked was fully POSIX compliant.

Android is an operating system which does not implement a Unix-like experience but does use the Linux kernel.

Systemd is a suite of applications including an init system which are designed to run as part of an operating system with a Linux kernel. In practise I have only seen it running on GNU/Linux systems.

Runit is an init system which is designed to run on Unix-like systems including GNU/Linux, Solaris & Mac OS X.

I have stated these facts to help expunge any confusion about these things which people might be experiencing.

[blackiwid]

I hope they will not sacrifice functionality and ease of use (for app developers and end-users, not sysadmins) in order to avoid dependencies. Whatever it will be, it's a suite of network configuration tools so by definition it *will* be Linux specific - and besides, I really think it's time to consider Linux as Linux, not like "one of the Unixes", so Linux-specific software doesn't worry me a bit.

I also don't worry about it, but I can say it's good that something else exist that is not linux specific. Free Sofware is about well freedom but also choice.

It's a bit like saying evolution should only been alowed on the Dinosouriers, and they are so successful let's ignore that mamals, they are pretty pathetic compared to the dominating Dinosauriers so they will never rise.

I also heard 10 years ago that Python is a bad language and that it never got used in big projects and it will never be... now it slowly becomes the Nr.1 Programming language (biggest grow rates).

As for Scheme-based service managers etc., I'm not fond of the idea. A service manager is one of the most privileges processes in the entire system, and it should be as static and declarative as possible. Having instead a full-blown programming language interpreter processing config in the form of executable code, thus prone to bugs, security issues etc., all the while requiring a garbage collector, is really BAD(tm) in my opinion. The fact that old-style init was effectively like that doesn't mean it was ever a good thing.

C as language has big security problems so the opposite of what you are saying is true. But I might be wrong. At the moment it is very secure because of its very small market share nobody would care to hack it except for maybe on very high priority target.

As long as NSA Rootkits are build into all our motherboards (except some GlugGlug users) I am not too woried about security.

The point is that this has to be tried out. A physics scientist also don't have said Newton works so great let's try to make that work as good as possible and forbid all other science on other stuff.

Maybe not the best analogy but close enough, but all the conformity stuff failed in the past big time we had LSB which defined RPM as the standard format in reality nobody installed any rpms on their debian machines. Heck it even got worse and worse over time now every programming language comes with their own package manager.

So we have created Virtualization and Container technologies to get rid of that problems and others, therefor this need of comfortably is bullshit.

I kind of believed something similar too in the past, I thought this distros outside of debian/ubuntu all suck and they do so much wrong.

Then I had a big awakening when Ubuntu tried to force me to stop using gnome (having very old versions and so on) then I mostly because of gnome tried Archlinux and Fedora and I was shocked how technically behind Ubuntu was back then. (before that I used gentoo but I became tired from 24/7 compiling)

Systemd is 1000x better (more user-friendly) than Upstart, it had current gnome, heck even the thing that Ubuntu did not have any way to get rid of the hundreds of old kernels installed on the system while these other distros deinstalled all but the 2 newest (configurable).

And I installed tools like profile-sync-manager that ubuntu if at all had only in a ppa and did not advertise that at all.

So you need to let things go. The motivation is important, a distro that just creates a alternative to systemd because some person disputes or some weird confusion what "unix philosophy" really means, will fail, but a broader concept that is more than only a reimplementation of something very similar will has good chances to become a big new thing.

You could argue that GuixSD is just a Nixos Clone, but it is different enough to probably make much sense.

I just wish they would include lvm support than I could finally use it

Also I learned with Emacs was Free Software or Opensource really means, before I used that it was more a theoretical concept that in theory I could get all sources and compile but in reality I nearly never did so.

With Emacs I feel like I can chance every behavior of my daily interacting with an eyeblink (or 2 3 shortcuts), so having systems that are easy fast instant hackable is a good thing.

Systemd was a similar move it made it easier for sysadmin to do work on such tasks, GuixSD and Nixos are operation Systems for Developers.

I don't care to much about the Service Manager, but I care about the configuration language and Nixos Configuration/Package language is not very good imho, so I would prefer such a scheme based version of Nixos, in the end it's about entry level, I must have really much pain that I start to work on C code, that is just a big pain in the ass, so I would likely never send any patch or do any patch on systemd, but if you have such a system with scheme it would empower me to do changes on that level, too.

Btw Scheme is not that high level the runtimes or compile times are pretty similar to C sometimes better. Well so far for now...

[jacob]

I also don't worry about it, but I can say it's good that something else exist that is not linux specific. Free Sofware is about well freedom but also choice.

It's a bit like saying evolution should only been alowed on the Dinosouriers, and they are so successful let's ignore that mamals, they are pretty pathetic compared to the dominating Dinosauriers so they will never rise.

I also heard 10 years ago that Python is a bad language and that it never got used in big projects and it will never be... now it slowly becomes the Nr.1 Programming language (biggest grow rates).




C as language has big security problems so the opposite of what you are saying is true. But I might be wrong. At the moment it is very secure because of its very small market share nobody would care to hack it except for maybe on very high priority target.

As long as NSA Rootkits are build into all our motherboards (except some GlugGlug users) I am not too woried about security.

The point is that this has to be tried out. A physics scientist also don't have said Newton works so great let's try to make that work as good as possible and forbid all other science on other stuff.

Maybe not the best analogy but close enough, but all the conformity stuff failed in the past big time we had LSB which defined RPM as the standard format in reality nobody installed any rpms on their debian machines. Heck it even got worse and worse over time now every programming language comes with their own package manager.

So we have created Virtualization and Container technologies to get rid of that problems and others, therefor this need of comfortably is bullshit.

I kind of believed something similar too in the past, I thought this distros outside of debian/ubuntu all suck and they do so much wrong.

Then I had a big awakening when Ubuntu tried to force me to stop using gnome (having very old versions and so on) then I mostly because of gnome tried Archlinux and Fedora and I was shocked how technically behind Ubuntu was back then. (before that I used gentoo but I became tired from 24/7 compiling)

Systemd is 1000x better (more user-friendly) than Upstart, it had current gnome, heck even the thing that Ubuntu did not have any way to get rid of the hundreds of old kernels installed on the system while these other distros deinstalled all but the 2 newest (configurable).

And I installed tools like profile-sync-manager that ubuntu if at all had only in a ppa and did not advertise that at all.

So you need to let things go. The motivation is important, a distro that just creates a alternative to systemd because some person disputes or some weird confusion what "unix philosophy" really means, will fail, but a broader concept that is more than only a reimplementation of something very similar will has good chances to become a big new thing.

You could argue that GuixSD is just a Nixos Clone, but it is different enough to probably make much sense.

I just wish they would include lvm support than I could finally use it

Also I learned with Emacs was Free Software or Opensource really means, before I used that it was more a theoretical concept that in theory I could get all sources and compile but in reality I nearly never did so.

With Emacs I feel like I can chance every behavior of my daily interacting with an eyeblink (or 2 3 shortcuts), so having systems that are easy fast instant hackable is a good thing.

Systemd was a similar move it made it easier for sysadmin to do work on such tasks, GuixSD and Nixos are operation Systems for Developers.

I don't care to much about the Service Manager, but I care about the configuration language and Nixos Configuration/Package language is not very good imho, so I would prefer such a scheme based version of Nixos, in the end it's about entry level, I must have really much pain that I start to work on C code, that is just a big pain in the ass, so I would likely never send any patch or do any patch on systemd, but if you have such a system with scheme it would empower me to do changes on that level, too.

Btw Scheme is not that high level the runtimes or compile times are pretty similar to C sometimes better. Well so far for now...

I don't think that what this is about. The main point, IMHO, is that service should be defined declaratively, not using a programming language. There should be One True way to launch, stop, suspend and restart services, express their dependencies and control their status. Any logic a service may require should be implemented using a service manager's API, not through an ad-hoc script. So yes, I like Scheme too, but the problem is not about switching to a better language, the problem is the need to eliminate any executable code from the picture.

[blackiwid]

I don't think that what this is about. The main point, IMHO, is that service should be defined declaratively, not using a programming language. There should be One True way to launch, stop, suspend and restart services, express their dependencies and control their status. Any logic a service may require should be implemented using a service manager's API, not through an ad-hoc script. So yes, I like Scheme too, but the problem is not about switching to a better language, the problem is the need to eliminate any executable code from the picture.

First of all there is a case to be made to have everything in one language. English is so successful because it's a standard many people in many sectors have agreed to as common standard, if we would all communicate in 20 50 different languages and everybody would need to learn 10 languages to be part a IT Professional as example the amount of people would be 1/100. of what we have today and it would slow down progress extremely. So having everything in guile/scheme isn't a bad idea.

Second "code" and not code is just a stupid differenciation if I have
runlevel = 5
or
(setq runlevel = 5)

is just a question if I have a parser between the markup "programming" language or if I directly use a real good programming language. Of course you could argue that with that separation you can define only a small part of code to be allowed / parsed, but you can do that with real programming languages too.

In a traditional distribution you have 1 mio config files all over the place and you spend 90% of the time searching that files. Putting 1000 of the small ones in 1 file alone is very beneficial, text editor is good in navigating in one big file, but they suck in switching between 1000 small files spread all over the place.

Having such config files is one abstraction that will make you unflexible, if you could make a improvement you will not do it because you wanna keep compatibility or it's harder to do that in a abstracted way.

It has on the other hand advantages, non-programmers or non-professional sysadmins (that can code a bit) have a easier time with it, but for the normal mass user will not be able to edit a service text file anyway, heck I try to avoid it as much as possible with 25 years linux experience.

One of the things I liked most about systemd, was that it is a simple enough system that maintainers and distros get easy into it and write such service files, before that some services you needed to manually copy some strange init scripts from homepages, I think as example for a BitTorrent client I was supposed to do that.

I also see that this is not for everybody, I could understand that you would be pissed if your maybe relative major distro uses it. But if a distro that is extremely different than what 90-99% of linux users use today anyway with a total different concept use it I can't see how that would hurt somebody.

This distros have NO STATE, that is very powerful, you usually don't write your own service files, if you have to do, something went wrong already, or you have a very special corner case.

So the Distro maintainers that code day in and day out in Scheme are the ones that should provide the service files.

Btw it's not like scripts do 1. then 2. then 3. it's more like service files "register service xy" with setting foo bar, now deal with it.

You don't have to code the start and stop procedure every time

It's hard to explain there are 2 different topics here shephard as a daemon and guix / nixos. In this distro you never edit any systemfile directly they are read only, so for all i matters the actuall code could be chinese code and written in Pascall or smalltalk.

It's functional distributions, that let you define your system more or less in 1 config file, and then you have a magic command that automatically transitions the system from old status to what you have defined.

That has many advantages, first you can keep the old systems and revert back to them or boot into them, 2nd as long as you have this config file, you can build this server identical or with a small chance (hostname as example on another machine) makes system backups not important anymore, you have to backup your data and /home and the configuration file but not your system.

But let's asume somebody uses shephard directly in a non-functional distribution:

Systemd Service file

https://patrakov.blogspot.com/2011/01/writing-systemd-service-files.html

vs Shephard decleration:

https://www.gnu.org/software/shepherd/manual/html_node/Service-Examples.html#Service-Examples

I am not sure if the "stop" is mandatory or optional but it looks pretty similar to me.

You can say shephard you need to write code to enable a service at boot time as it looks. But some people are more happy to edit a file than use a console with some commands.

So that's not bad or good it's just a different taste.

ok have to take that back seems to be only a optional thing, I don't know:

Actions that are available for every service are start, stop, restart, status, enable, disable, and doc.

So to make it short it's not a perl or shell script that does all the work, you don't have to write code to get "status" or if it crashes it probably will restart it. I don't see a big difference therefor between the two except different syntax of cornfiguration. It's also a deamon runing outside that does the work, the service itself is no program, you still only register it.