systemd 250 Released With A Huge Number Of New Features, Improvements

[coder]

Joke not accepted; got it.

pal666 only seems to like jokes at others' expense.

Given how much that user rips on everyone else, being a good sport about taking some in kind is the least they could do.

[coder]

I've rolled my own distro for a decade.

Please post a link.

What other things do you do, besides that? Do you have a full-time job? Is any of your paid time spent maintaining the distro? Do you have a family life? Friends? Are you civically-engaged in your community?

Just because you're willing to maintain a distro, which I'd wager probably uses entirely off-the-shelf components written by others (i.e. not your own init system) is no answer to anyone's complaints about systemd.

[jacob]

Do you allow anyone who watches movies to have an opinion about them, except other film makers? What about allowing consumers to have opinions about products they buy? It's the same thing. I'm allowed to have an opinions about what I deem is a better solution than the one I'm currently stuck with. That does not qualify as a sense of entitlement.

What I don't understand is why you're apparently so threatened by any form of criticism. Why are you so invested in systemd, exactly how it is? I justified my compaints. Now it's your turn to explain your defensiveness.

I'm not invested directly in systemd except in a general sense, as in it's something that brings real improvement to the FOSS ecosystem. Your analogy doesn't hold water. If you don't like a particular film director/producer's movies, don't watch them; if you don't like cars of brand XYZ, buy some other one. This is different. The anti-systemd crowd could similarly just use their Devuan or Void or whatever and accept that they are off the mainstream with all that it implies. But no, instead they behave as if Linux and FOSS owed to them and they had the right to decide what is legitimate and what isn't. Yet of course they usually don't contribute anything, especially not to upstream.

Secondly, you haven't explained your criticisms except that for some unspecified reason a driver database "should not" be maintained under the systemd umbrella because you say so. But that's neither here nor there. You haven't justified why in terms of better outcomes (sorry, "unix philosophy" doesn't cut it as an argument) and you haven't said who would be doing it, how and why. On the other hand here is a team that spends time, energy and resources doing exactly that and all your criticism essentially boils down to that they should make life more complicated for themselves as well as for 95% of users, only to accommodate the wishes of a tiny, non-contributive but aggressive minority. If that is not a sense of entitlement then I don't know what is.

[oiaohm]

A Linux distribution worthy of the name must, in addition to being accessible to the public, be maintained by security updates. Only one person cannot maintain a distribution and provide it with the necessary security.

The reality here is in the embedded world there are about 100 new distributions of Linux made per day with almost 100% of them coming unmaintained from day one of the product release. They were maintained in the development cycle of over 1 year with all necessary security updates by normally 1 person.

So its possible for one person to make a secure Linux distribution but is not wise long term in most cases.

What exactly are you talking about? Worthy of the name? I argued that you can build your own distribution, and you can! It covers my security needs 100% and I'm using it as a daily driver for my HTPC, and I've also ported it to several ARM platforms (it even earned me some freebies from the manufacturer) . It's available on github (albeit an older version, I don't publish my updates anymore due to time constraint)..

Yes time constraints is one problem https://en.wikipedia.org/wiki/Bus_factor the bus factor is another. This custom distribution of your is not exactly ideal lets say you get sick and cannot maintain it for some reason of course while you are sick you will be wanting to use your HTPC right.

I would say a single person can make a Linux distribution from scratch today. But its absolutely not recommended. Yes time cost and bus factor is what makes it not recommended. Do remember a lot people do not recommended things.

There are a lot of examples of so called custom forks of existing distributions end up being user torture devices due to the dispute between the custom distributions added repositories and the base distributions repository. So making your own custom ISO with your own custom package added parts to existing distribution has the time constraints problem and the bus factor problem and the I am going to be unstable problem. Yes those making docker images learn this one.

Warped right building a complete Linux distribution from scratch end up more stable and less painful than extending existing in a lot cases but the up from cost is very prohibitive..

There is a reason why we have these days flatpak snap docker and systemd portable services.

[coder]

I don't publish my updates anymore due to time constraint).

there is absolutely nothing stopping you from creating your own distribution completely from scratch if you will.

except your above point.

Anyway, it's a false solution. It's like saying that if you didn't like a how a movie ended, nothing is stopping you from taking your phone and shooting your own movie. Technically true, but not really a practical alternative for the big budget movie just not screwing up the ending. You can't get the big-budget effects or quality actors & soundtrack composers. It's the same thing with systemd. I can probably install some second-rate components that would yield a working system, but they don't have feature-parity to systemd due to systemd using up all the oxygen.

It's exactly the same thing as the Linux kernel. It's far from perfect, but most of us run it because it's the best option currently available, due to being the entrenched solution and benefiting from the Network Effect. The effort to replace it is virtually insurmountable, as Google seems to be learning, and even then you won't have the same level of drivers and hardware support.

[coder]

You can't just "add ioctls" if the hardware lacks the capability to self-describe.

Obviously. However, many devices can. So, unless an entire device class fundamentally can't self-describe, then you could implement & use them for those devices which can.

And even in the opposite case, it needs a ton of effort in every driver that no one is really willing to spend.

Really? A ton of effort?

I say just add the hook and that way the hwdb needn't contain any defaults or overrides for whichever devices can. Over time, this should become an ever-greater proportion.

It's a bit of a contrived case. I don't think it applies to anything except SCSI/ATA storage devices, and udev is already doing that (see how block device probing works).

For all I know, it could easily apply to certain classes of USB devices, newer revs of HDMI & DisplayPort, CXL, DDR5, etc. Newer connectivity standards tend to be much more self-descriptive and richer in the queries they support.

[oiaohm]

The core systemd developers are paid employees of Redhat tasked primarily with its development. I have different specializations & priorities, and so does my employer. We use systemd not by choice, but by default. While I believe a truly modular solution would be better, the pain experienced by systemd's shortcomings doesn't exceed the activation energy needed to actively pursue a different solution. This doesn't nullify any of my points. It would take "a lack of technical competence and understanding" not to comprehend that.

That the reality. Lot of people want a modular solution of course there was not the activation energy to maintain that. Yes the reason why you don't have the effort to go after something else. There were a lot of short comings in the modular solutions from the time before systemd existence.

Systemd is quite modular when you get into custom building it.

https://elinux.org/images/6/69/Demystifying_Systemd.pdf

Yes this is 2016 yet people today in embedded still make cut down versions of systemd. Yes it possible to run systemd without udev or journal.

The Soletta project noted in that PDF fairly much died.

Please note I am not saying systemd is without it short comings but its really easy to miss the historic short coming of the modular prior to systemd.

Having sysvinit, udev, consolekit... as independent projects to be fully modular resulted in a lots and lots of work to make sure that they would work with each other. Yes you would hit stupidites of patches merged in the wrong order. So you could have a patch in sysvinit that broken udev yet was required for consolekit. Yes of course the fix for udev would be sitting in the mailing list rejected due to some coding error.

The we had sysvinit having patches not going mainline for 3 years because the maintainer was no longer live. (yes I am serous person 6 feet under) and no one step forwards to replace the person or worked out that was the problem.

Modular has it own fair share of problem. Big one is really simple to end up with one part without enough maintainer people to the point that people are not quite picking up that a serous problem has happened. Second one is integration testing to make sure that the complete solution can work so you are not adding patches to one part that break others because those other parts have not got patches yet.

coder like it or not history of sysvinit and it modular setup tells us we did not have the activation energy to properly maintain a modular setup and nothing has changes to suggest it was attempted now that it would turn out any better.

[coder]

I'm not invested directly in systemd except in a general sense, as in it's something that brings real improvement to the FOSS ecosystem.

I'm not even saying it doesn't.

Your analogy doesn't hold water. If you don't like a particular film director/producer's movies, don't watch them; if you don't like cars of brand XYZ, buy some other one. This is different.

Virtually every car owner you ask is probably going to have at least a couple gripes about their car. This will be true of any complex product or system. You're saying we're not allowed criticize the short-comings of a solution, as long as there are alternatives? That's madness, unless maybe if you live in like North Korea.

At this point, your position is so blatantly unreasonable I probably don't need to say another word.

The anti-systemd crowd

Ah, so if I don't love everything about systemd, then I must be a hater?

could similarly just use their Devuan or Void or whatever and accept that they are off the mainstream with all that it implies.

Ah yes, because haters don't deserve anything good. I see what you did there.

But no, instead they behave as if Linux and FOSS owed to them

They have as legitimate an opinion and as much at stake as everyone else. If Debian hadn't gone with systemd, the pro-systemd crowd would be just as bitter. You can't see this, because you have the illusion of moral righteousness on your side. You don't allow for any possible realities except the present. You can only see what we had before systemd, and presume no progress would've happened had systemd not won out.

and they had the right to decide what is legitimate and what isn't.

Exactly as you're doing.

Yet of course they usually don't contribute anything, especially not to upstream.

Most people with a stake don't contribute, because it takes a lot of resources. It's exactly how people in a democracy don't all run for political office, or how consumers in a capitalist society don't all found companies building products to compete with ones they deem to be sub-optimal.

You're setting the bar for criticism much too high, and then using this to demote those who dare to speak out to some second-class status, where their concerns become less valid.

Secondly, you haven't explained your criticisms

See post #67.

you haven't said who would be doing it, how and why.

I've spent enough time debunking this fallacy. Criticism need not be buttressed by the existence-proof of a counterexample (nor even a concrete counterplan with committed resources).

they should make life more complicated for themselves as well as for 95% of users,

First, I dispute that it would actually be more complicated, overall. If there would be fewer bugs, that would be one net-positive. Second, if competition yielded better solutions to certain problems, or a variety of solutions which are better tailored to specific usage models, that could be another advantage. It's actually harder for you to argue against the counter-factuals.

Second, I dispute that you represent the interests of 95% of users. You may speak for yourself, and perhaps hard-core systemd adherents. Possibly even most of us who use it probably don't think it's optimal, but merely the best of what options they currently have available.

a tiny, non-contributive but aggressive minority.

First, you already acknowledged there are non-systemd alternatives. So, the non-contributive part is false.

Second, anyone who's disenfranchised is naturally going to be aggressive. Again, this would be true of the systemd people, had it been rejected.

Third, the fact that there are more contributions to systemd doesn't mean it's perfect. As the established player, it gets contributions by default.

If that is not a sense of entitlement then I don't know what is.

If you want to know entitlement, just listen to yourself. You're so entitled that you think your preferred solution is beyond critique, simply because it's dominant. You seem to confuse dominance with perfection.

For some time, Microsoft Windows was the dominant OS. That didn't automatically mean those who decried it were entitled computer users. Even those who weren't active contributors to another OS.

[arQon]

Using "excessive mounts" is the only way to do a lot of things in Linux. That's simply how this operating system works.

Yeah... except that prior to systemd, we somehow managed for 20+ years without "having to" do that.

And again, you're simply fanboying - you have no idea what you're talking about, but you feel obliged to defend systemd from any criticism at all, because it's a religion for you or something.
Fun fact: it isn't actually the law that you HAVE to be an asshole just because this is the internet, despite what you seem to believe.

[coder]

like it or not history of sysvinit and it modular setup tells us we did not have the activation energy to properly maintain a modular setup and nothing has changes to suggest it was attempted now that it would turn out any better.

That's a fair point, but it's also not proof that modular approaches aren't viable. First, SYS-V Init was far from perfect. An improved replacement could be less fragile, resulting in fewer dependency issues like you mentioned.

Second, that's actually going back pretty far. The Linux world is a lot bigger, now. The ways and places it's being used are a little more diverse, with a lot more resources being invested in each.

Third, at the time, there weren't modern, stable, well-specified interfaces for most of the modules. That's what you need for a properly modular approach to be viable and sustainable.

For the time being, we can't know how the world would look if systemd were more narrowly-focused (or such an alternative had won out, instead), and independent solutions were devised for other parts of the userspace services it now encompasses. So, I'm less concerned with arguing these counter-factuals and more just wishing the systemd devs would seriously consider not continuing its apparent Manifest Destiny.