[mdlug] Big brother gives M$ a 10 out of 10

[Ron / BCLUG]

LAP wrote on 2024-05-24 04:34:

>> On May 23, 2024, at 12:25, Steve Litt<slitt at troubleshooters.com> 
>> wrote:
>> 
>> 
>> I’m curious, how does systemd threaten that OS model?
> 
> Let me first state that I am passionately against systemd

Sometimes a *dispassionate* analysis is enlightening.



> Systemd is still very much a work in progress

Isn't everything a work in progress? Linus is still involved in patches
to the kernel ~30 years on.

systemd is a ~14 year old concept and has been included in distos for 
~13 years now. It's fairly mature.



> but the project definitely has a goal which is stated in a talk given
> by Poettering some 10 years ago.  Everyone should read his simple
> manifesto:
> 
> http://0pointer.de/public/gnomeasia2014.pdf

It's 85 pages, and 10 years old; I just skimmed it.


> The goal of systemd is to become the one-and-only interface between 
> user space and the kernel.

Isn't that just "standardization"? It doesn't sound too bad, especially
for developers of Linux software.


> Poettering refers to the current state of Linux as being a "bag of 
> bits."

I saw /etc/default vs /etc/sysconfig as one example of "pointless
differences" mentioned in the link you provided. Seems valid.

These accumulate and can cause needless difficulties for developers and
maintainers and have no benefit to anyone.


> This is definitely a serious condemnation but it is Poettering's 
> basis for his desire to transform Linux into a "competitive" OS -- in
> purely his image, of course.

Poettering is unpopular, but don't let that interfere with a
dispassionate evaluation of his ideas.

He's not forcing this on Canonical, for example. They see something
worthwhile there and adopted systemd over their own Upstart.




> But being a "bag of bits" is actually the great strength of GNU/Linux
> because it really means that there are always many different ways of
> doing most anything.  Linux offers a plethora of virtual terminals,
> graphical toolkits, scripting shells, init systems, etc., etc.  This
> great bounty, however, is upsetting to the vision of unity that
> systemd aims to impose.

It's great to have varying ways to accomplish something, but if they're
not portable between, say, RHEL and Ubuntu, that's a problem.


> From the beginning I have rejected systemd.

Would the fact that it's become the common and accepted way of doing
things - and that it works rather well - change that, or will you 
continue to reject it no matter what?


I'm reminded of Stephen Colbert at a White House Correspondents dinner:
"We all know that George W Bush will believe on Wednesday what he
believed on Monday.

No matter what happens on Tuesday."



> It is present on none of my GNU/Linux machines.  But it is getting 
> more and more difficult to avoid and the great threat is that it may
>  impossible to avoid in the future.

There's always BSD.


> As an example, since I build my own machines from components, I know 
> exactly what hardware is present in my system and consequently I can 
> create a convenient set of static device nodes.  There is no need to 
> probe hardware at each boot (indeed, it would be a stupid exercise). 
> However, freedesktop.org, which is actually a part of the systemd 
> movement, introduced libinput as the one-and-only input driver for X
>  Window.  Libinput does not recognize static device nodes.  Rather it
>  uses udevd (from systemd)to probe hardware and then create device 
> nodes "on the fly."

Your usage case is not common - Linux needs to handle transient devices
(monitors, network interfaces, storage devices, ...) or else it'll
wither and die.

It's unrealistic to expect an edge case to be catered to at the expense 
of 99% of modern computing usage scenarios.



Also, can I call out the fact that Xorg / X11 has its own text editor
(xedit) and a print server
(https://www.x.org/archive/X11R6.9.0/doc/html/Xprt.1.html)?

Talk about bloat! But, stuff like that never gets mentioned by 
anti-systemd folks.





> Thus, if it is not plain already, the threat of systemd is that it 
> will destroy Linux diversity and institute a completely monolithic 
> set of utilities under the control of one organization.

a) s/threat/promise/ re: institute common utilities
b) monolithic is inaccurate - it is a collection of single task tools
c) "under control of one organization" can be rectified via forking, 
right? It's open source. https://github.com/systemd/systemd


> Others will certainly disagree, and there are obviously a lot of 
> these others as most GNU/Linux distros have adopted systemd.

Yes. They (Canonical, Suse, RedHat) adopted Upstart, then switched to 
systemd.

Have you ever wondered *why* there was an interest in developing a
services management system common to all Linux distros?

That's crucial to a dispassionate analysis of what's going on.

Once one recognizes multiple efforts launched to do that, one must
consider the reason.

Then one can ask if SysVinit > systemd > Upstart.

Seems even Canonical agrees with that. They're famous for attempting
their own solutions and rejecting "not invented here" software (Upstart,
Mir, Ubuntu phone / touch, Unity,...)




> The crux of the matter is that systemd should be a fork and not an 
> imposition.

That's an opinion that isn't shared by those that do the heavy lifting
in creating distros and software development.


> There should be a divergence of methodologies and not a convergence.

There needs to be a common method of doing things - a convergence.

No one wants to maintain multiple branches of $linux_app for the RHEL
and Ubuntu and Suse methods of doing things.



> If one reads the manifesto, Poettering mentions a "competitive" OS 
> goal for Linux.  I would hope that everyone can see that what this 
> really means is that Linux must become more like Microsoft Windows.

Nonsense, systemd does not make Linux more like Windows.


> Indeed most of the goals of systemd are a direct copy or elaboration
>  of Microsoft capabilities.

Windows and Mac supposedly both have services management systems.
systemd is most often seen as taking after Mac's launchd, not Windows.


> I do not use the common DEs such as GNOME or KDE, but from what I 
> gather, these DEs are not much more than MS Windows GUI clones.

That's a real stretch unless one considers WIMP (Windows, Icons, Menus, 
Programs) framework as a "Windows clone".

Most people consider it a standardized functional and intuitive 
interface design settled upon decades ago.

Gnome looks quite different (I don't use it, cannot be sure) and KDE 
kind of implements it, but is so configurable that it can be made to do 
/ look like anything.



> In conclusion, systemd is still a work in progress.

Everything is a work in progress, unless one thinks like the Amish 
(technology in eighteen dippity doo was perfect, no more advances).



> Unfortunately, a lot of users falsely believe that systemd is merely
> an alternative init system, but this is way off the mark.

Agreed!


> Systemd is a movement to comprehensively restructure Linux and as a 
> result destroy the Unix way of doing things.

Yes to restructure, no to "destroy the Unix way".



> It could very well be that the future of GNU/Linux, as a free and 
> open alternative OS, will be FreeBSD.

I'm unaware of any surge in BSD installations and corresponding drop in 
Linux installations in the years since systemd became common place.


It's almost as though people tried it, evaluated the pros & cons, and
just carried on unbothered.


rb