Devuan Jessie beta released

Posted May 1, 2016 19:46 UTC (Sun) by darwish (guest, #102479)
In reply to: Devuan Jessie beta released by Zack
Parent article: Devuan Jessie beta released

anselm provided concrete systemd features that I've personally seen used everyday on production server _and_ embedded systems. And now what was your reply? Striking all of these features down just because they were done using a kernel feature that is not to your liking? And what to do you really propose in place? I guess you'll need to study Neil Brown's series on cgroups here first: https://lwn.net/Articles/604609/

> "a number of separate pieces that only work together in limited ways?" That's a nice way to summarise Software Engineering 101. If only some people would paid more attention to it in class instead of writing unspecced evergrowing solutions in search of a problem.

A number of separate pieces that communicate by text over dumb pipes does not handle the computing challenge of this day and age.. This method has its uses (e.g. quick scriping, tinkering, & administration), sure, but please let go of the 1970s lingua franca of marketing this design pattern as the only solution to all the problems of computing .. Have you looked at GNU autotools 10K+ scripts? They are damn ugly and untracable ..

And no, no .. software engineering classes are not taught with this "trivial tools connected with pipes" design pattern. Almost all of the software you're using right now, from the LWN.net python code written to the billions of iOS and Android devices in the hands of users are written using the concepts of communication buses and object oriented interfaces (just like the systemd dbus interface, btw, which helps systemd being __properly__ integrated in much bigger solutions) .. sorry to break your bubble ..

to post comments

Devuan Jessie beta released

Posted May 2, 2016 4:10 UTC (Mon) by Zack (guest, #37335) [Link] (5 responses)

> that is not to your liking?

or any serious kernel hackers', but let's reduce it to just one person, because, I don't know? Why do the vocal people who like systemd always do that?

> And what to do you really propose in place?

In order of preference:

1) Something that does what cgroups does, sanely? You know, build quality software on top of quality software? Not starting to build the house before you drained the swamp? Then again, castles in the sky don't really need foundations.

2) Nothing

3) cgroups, but let's not base an important part of infrastructure on it because that would be a really silly idea that, frankly, should be quite self-evident.

> but please let go of the 1970s lingua franca of marketing this design pattern as the only solution to all the problems of computing

Systemd certainly introduced a whole new school of thinking: "Those who do not understand how to poorly reinvent Unix, are condemned to reinvent Plan9, poorly."

> I guess you'll need to study Neil Brown's series on cgroups here first: https://lwn.net/Articles/604609/

A series of articles on how to use it, since it's there anyway, which it shouldn't be, but systemd cemented it there, and now it won't go away.

Now let me in turn quote an objective analysis that refers to the actual quality of cgroups: "What the fuck? The interface in question is optional; it's *NOT* guaranteed to be present in any kernel. Disbelieve all you want, but that fact does depend upon your beliefs. Incidentally, the code behind that interface is a misdesigned piece of shit, best configured out unless there are extremely strong reasons to enable it. Not everything that can be enabled should be."

> "trivial tools connected with pipes"

No, "separate pieces that only work together in limited ways", as opposed to "a big ball of mud that works together in unlimited ways, all alike" but feel free to reduce it ad absurdum some more. Why do you do that again?

> the billions of iOS and Android devices in the hands of users are written using the concepts of communication buses and object oriented interfaces.

Ad numerum, but even if it wasn't, that's actually a pretty decent counter argument to why it would be a good idea.

Devuan Jessie beta released

Posted May 2, 2016 6:26 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link] (2 responses)

> 1) Something that does what cgroups does, sanely?
That "something" will look almost exactly like cgroups.

> You know, build quality software on top of quality software? Not starting to build the house before you drained the swamp? Then again, castles in the sky don't really need foundations.
Core cgroups have acceptable quality and systemd doesn't need anything more. cgroup controllers are NOT required by systemd at all.

> Now let me in turn quote an objective analysis
No, it's not. Since you obviously have no idea what you're talking about.

> that refers to the actual quality of cgroups: What the fuck? The interface in question is optional; it's *NOT* guaranteed to be present in any kernel.
So is pretty much everything else, including even filesystems. A typical distro depends on multiple optional kernel features.

Devuan Jessie beta released

Posted May 2, 2016 14:05 UTC (Mon) by Zack (guest, #37335) [Link] (1 responses)

> Core cgroups have acceptable quality

Maybe for someone who runs OSX on his desktop/laptop and solely uses containers for deploying software on the servers at his workplace and is firmly opposed to how distributions package software as integrated packages.

I wouldn't readily accept any definition of "acceptable" or "quality," or "acceptable quality" from such a person at face value though.

> No, it's not.

Okay, you got me. It's not, strictly speaking, an objective evaluation, but it's also not mine. It is the evaluation of someone who can be trusted to make an evaluation as close to objective and as far away from politics as is possible and definitely has an idea what he's talking about.

Devuan Jessie beta released

Posted May 2, 2016 17:49 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

> I wouldn't readily accept any definition of "acceptable" or "quality," or "acceptable quality" from such a person at face value though.
So no technical arguments against my statement that systemd doesn't depend on 'bad functionality' of cgroups?

OK.

Devuan Jessie beta released

Posted May 2, 2016 15:51 UTC (Mon) by Doogie (guest, #59626) [Link]

>> that is not to your liking?

>or any serious kernel hackers

Nor any true scotsmans!

Devuan Jessie beta released

Posted May 5, 2016 13:34 UTC (Thu) by glaubitz (subscriber, #96452) [Link]

> or any serious kernel hackers', but let's reduce it to just one person, because, I don't know?

I'm sure you have a source for that you can quote, don't you?

Adrian