Using Udev Without Systemd Is Going To Become Harder

[valeriodean]

Exactly. I actually like Lennart, and we are in good terms. I'm not happy that this particular thread post was referenced in this Phoronix post in this way, because it doesn't give
an accurate picture of what is really going on.
I wouldn't mind systemd becoming the Gentoo default, as I see systemd becoming the norm in Linux userspace, but that hasn't happened yet, and migration will need time.
So, with all the best in mind, for everyone, for everything, I was only expressing that I'd like udev from systemd tree be usable without *an running systemd instance*
for longer time, so users don't get too dramatical change too fast.

Thanks for sharing your opinion, but the purpose of this article was to start another flame war around Lennart and systemd. Sorry but your opinion will disappear in the middle of the fight.

[gens]

I have loads and loads of times. Linux is not about choice. You have choice in e.g. linux kernel, that's what you shown. Now prove that Linux is _about_ choice. Huge difference! Saying Linux is about choice means that choice is a goal in itself. It's not.

yes it is, the kernel especially

you can compile the kernel without the whole network stack and it would work np

linux is about choice, and the kernel's goal is to be flexible and NOT assume anything
(it is almost a perfect unix clone, and unix is MADE to enable people to do whatever they wanted without it getting in the way)

only thing where you have no choice is X, and even then things like SDL can work without it
(that is until systemd came, but you still can get everything except gnome3 without it)

[bkor]

Completely off topic, but I disagree.
Modular makes designing and building harder, but maintenance easier.

That's obvious in the very example you choose: It is indeed harder to assemble a poster of 90 pieces than a poster of 9. But that is not maintenance, it's building. On the other hand, if you have to redraw a piece, it's obviously easier to redraw a piece that is 10 times smaller.

I disagree. How is it easier? Say you have a section that's 20% of the poster. You want to change it. It's either 1 piece or e.g. 10. I'm pretty sure making changed to 1 is quicker. Don't have to piece those small ones together. Meaning: all those interactions add complexity and are source for potential cause for bugs.

If you have a public API people will depend on it. You're restricted in what can be changed. Sometimes you have an error in your API and there's nothing you can do to really fix it, because you're breaking API stability. Read the comments regarding openssl API. It's something that you can use and terrible. Now try and fix this.

[bkor]

yes it is, the kernel especially

Can you give arguments other than "yes it is"? This feels like kindergarten!

even most of the user facing things are about choice
you can run wayland, X or just make your own over a framebuffer
in like enlightenment or xfce you can move the toolbars, move/place/replace/remove/resize buttons and widgets, change themes etc etc
in a console you can alias commands or replace them with whatever, change autocomplete behavior or turn it off completely, etc etc

As said before: Having choice is different than being _about_ choice. Wayland is not about choice, it is about doing things better. XFCE has options, it is not created to provide choice (it is a desktop environment).

you can compile the kernel without the whole network stack and it would work np

linux is about choice, and the kernel's goal is to be flexible and NOT assume anything
(it is almost a perfect unix clone, and unix is MADE to enable people to do whatever they wanted without it getting in the way)

Again: Having choice is different than saying choice is a goal in itself. Your reply doesn't address this at all.

[johnc]

To all the people saying "Linux is not about choice":

That was a GNOME developer, and believe me, GNOME people don't know anything about choice these days.


The problem with the idea that Linux should "be about choice" is that it threatens the market leader and the ability of the market leader to control its status as the market leader. We see, e.g., how Android is becoming less about choice as it has matured and Google wants to make sure its grips on that golden goose don't loosen. Conformity empowers the market leader. E.g., we're now hearing about how third-party skins are being discouraged in Android's future. In other words Samsung, know your place. And don't think about coming up with an alternative Android. And think less about trying to differentiate your product from ours.

Likewise with Linux. You want to have the entire OS consolidate around a single set of intertwined technologies that would prohibit alternatives from coming about without first requiring an enormous amount of resources, which, of course, nobody but the market leader has. In other words, eliminate differences and opportunities for differentiation. Let the only difference be the color of the icons on the desktop or something superficial or irrelevant such as that. When there is no differentiation of product, what need is there for the consumer to step away from the market leader?


The issue with Gentoo is that it is the base for Chrome OS, so you better believe the Market Leader of Enterprise Linux needs to get in there and but the kibosh on that shyt before things get out of hand. And Google, after all, has the clout to compete with the Market Leader of Enterprise Linux if it so chose to do so. It's okay to have some distro out there that's just the private tinkerings in somebody's basement, but once you get on the playing field and pose any kind of threat to the market leader, it's time to brandish the swords.


To the people saying, "Oh! If you don't like systemd, just make your own alternative!": it's clear that you are clueless. When making an alternative to systemd now requires a massive multi-million dollar investment, you've pretty much eliminated just about everybody's ability to make an alternative. And all you have done is allow the market leader to eliminate further competition.

Some will say, "Linux is FOSS, do whatever you want." But obviously a multi-billion dollar market-leading company has learned how to play the FOSS game to their own benefit. These are for-profit companies we're talking about here. They're not in it for the religion.

[JustADirtyLurker]

Embrace.

Extend.

Extinguish.

We are waiting for the next project to be swallowed and crippled by Lennart.

For better or worse, systemd and pulse are the reality of Linux for the next year or more. Once they implode, we'll need to move over to something else, as regularly happens in Linux. We've survived hal, devfs, esd, arts and a million other things, so let's hope that whatever succeeds systemd is going to be saner and managed by less destructive individuals.

I have to be honest. I haven't seen any breakthrough ramp up in ages at system level up until systemd arrived. systemd, which started as an init system, is becoming more and more a big companion for the kernel for providing userspace, services, devices (udev). regardless of the OP discussion (udev depending on systemd or not), I wouldn't bash the guy Poettering for his attempt to enhance the mess that there's at low level linux.

There's too much conservatorism in the linux world, as soon as somebody tries to touch something that has "fairly" worked for decades because it is believed that it can be improved, the hating brigade ramps up with all their claims (eg. see X vs. Wayland debate: "noooo, we lose network transparency!!"... d'uh). I firmly believe that we need to embrace enhancements actively: it makes no sense to stick with obsolete software just because the new one is not a *nix way of doing things...

If systemd is finally widespread this could also lead to, finally, have some standardization at the base OS level. Quite frankly, I'm tired of all the "choice" we have. It's not only the kernel's fault if we still have no stable APIs, no standardized interfaces, no binary compatibility.
I don't mean to say that Lennart and Kay Sievers have all the answers to this problem, in fact I'm quite happy with all the people keeping (trying to keep) them in line. We are talking about a very delicate piece of software after all.

[rob11311]

I hate Lennart's attitude, he acts just as a annoying spoiled kid
...
You see that when he presents a patch, he already includes the words "systemd haters". This way he can avoid any opposition. It's like some political minorities nowadays that call everyone who doesn't like anything they do just "haters".

Yeah! The arrogance shown is astonishing! Lennart does himself no favours.

To make this clear, we expect that systemd and kernels are updated in
lockstep. We explicitly do not support really old kernels with really
new systemd. So far we had the focus to support up to 2y old kernels
(which means 3.4 right now), but even that should be taken with a grain
of salt, as we already made clear that soon after kdbus is merged into
the kernel we'll probably make a hard requirement on it from the systemd
side.
...
Also note that at that point we intend to move udev onto kdbus as
transport, and get rid of the userspace-to-userspace netlink-based
tranport udev used so far. Unless the systemd-haters prepare another
kdbus userspace until then this will effectively also mean that we will
not support non-systemd systems with udev anymore starting at that
point. Gentoo folks, this is your wakeup call.

It's reasonable to say, feature X requires Y, but critical system software OUGHT not depend on stuff in a brittle way, it's just not how you build a robust system. He may not mean what he actually says, but if you're a developer who may need to test and develop, a patch to an old kernel, you're not going to feel very happy with Lennart's attitude.

It's a real shame as I like the core concept of systemd, I just am continually made anxious by the lack of apparent taste by it's main developers.

The X changes away for config a few years back, broke so much stuff for so many, because they didn't keep older chips / monitor handling stable, when they did re-write.. compatability takes care and effort, not this "I know best!" approach.

[rob11311]

When making an alternative to systemd now requires a massive multi-million dollar investment, you've pretty much eliminated just about everybody's ability to make an alternative. And all you have done is allow the market leader to eliminate further competition

What's really stupid about the approach shown, is Lennart is likely to fragment his own installed base if he couples things with the kernel too closely. You'll have kernel X from distro, System D Y and learn never to update 1 without other from same source, which will lead to 1000's of systemd versions and variants, likely to lead to misery. Especially if systemd breaks Linus's laptop.

[rob11311]

What a inaccuracies in one post. Kay isn't writing kdbus. Kdbus is written by Greg. Kdbus will now use other infrastructure not written by systemd people. Kdbus being reviewed is totally normal and is a good thing. You're pretending it is bad code being pushed without anyone looking. Nice usage of Fear there!

Look at this source - KDBUS Talk LWN article

GKH is described as "involved".. you don't seem to have followed the rather unusual events and drama, WHICH have conflated a number of different issues, with various key developers. Linus in particular linked GKH and Kay...

[johnc]

"Upstart": a term clearly selected because of its double-meaning, a fact certainly not lost on Red Hat.