The Politics of systemd Part 2

[Tony0945]

. People dont want to think. They dont want to learn. They just want to consume.

That's what Windoze is for. I think at least 50% of Linux users would go back if Windows were $39.99 instead of $300

[gwr]

all that, GPL provided, will back-feed into the rest of the ecosystem.

The future is bright.

Not with systemd it isn't. One of the main political factors with systemd is it's allowing proprietary software to bypass the GPL by linking to systemd RPC mechanisms. Once systemd becomes the main interface to the kernel, then there is nothing stopping proprietary tools.

viewtopic-p-7645524.html#7645524

[khayyam]

all that, GPL provided, will back-feed into the rest of the ecosystem. The future is bright.

Not with systemd it isn't. One of the main political factors with systemd is it's allowing proprietary software to bypass the GPL by linking to systemd RPC mechanisms. Once systemd becomes the main interface to the kernel, then there is nothing stopping proprietary tools.

madjestic, gwr, et al ... besides that, this ecosystem (like any viable ecosystem) has always consisted of diverse, and varied, flora ... it wasn't designed by the maker of OSes, it developed organically by virtue of principles internal to it (ie, "small tools doing one thing well", interchangeable with other tools, etc, etc). This is now described dismissively as a "bag of bits" by systemd developers (see the Gnome Asia 2014 presentation), so that its very strength is cast as a systemic weakness (the 'legacy' mantra). Not only does systemd undermine this ecosystem in terms of its "one true way", it is poisoning the well in terms of those things which are in fact its nature. So, no, the future isn't "bright", unless of course you're happy with the idea of a systemd monoculture, and all that implies.

best ... khay

[augustin]

The latest news about KDE and systemd is pretty saddening.

Is there a thread dedicated to KDE and systemd, around here?

Edit:
Ok, here is a new one:

KDE and systemd
viewtopic-t-1034190.html

[tld]

It feels like modern desktops are, to a large degree, about wrapping up low-level stuff in shiny GUIs, where presentation is valued above efficiency. What's worse - some desktops paradigms pretend they know better how a user should interact with the machine. It attracts users with certain mindsets and, together with developers of certain mindset and marketeers, it creates a feedback loop.

Years ago I used KDE and that was the feeling I always got. Worse yet, at the time, the few things I really wanted it to do never seemed to work right (I never did run a version that didn't constantly forget the application I assigned to file types). What I hated the most was that it was a complex black box where I never knew what God-forsaken, impossible to find, XML or config file (either in my home or some default somewhere else) was behind some behavior.

Then I switched to using nothing but fluxbox, where basically everything it does is driven by something in ~/.fluxbox that you can actually find. Maybe I'm an extreme case, but I discovered that I didn't even miss having desktop icons, nor any sort of file explorer program, and never installed anything to add them. I have fluxbox shortcut keys set up to launch the applications I use the most, and do most things from the command line in urxvt (I have a fair number of bash aliases) and have never looked back...and everything I do is easier and more efficient. I've been working like that for years now and wouldn't trade it for the world.

Tom

[evoweiss]

Hi all,

Just chiming in here on the topic. I cannot claim any competence or knowledge in the technical ins and outs of systemd. However, I'm a long time Gentoo user and I like its flexibility, choice, and everything else about it. I am not enamored of new and shiny things just because they're new and shiny. I like what works and works well. For most matters, I am pretty cautious about moving to something new. If it does the things I need better, sticks to principles that work, and is free of bugs, I am happy to migrate. Systemd is none of those things and represents a step backwards in my view. The fact that it makes it a pain to debug problems makes things worse. Should the time come (shudder) when even Gentoo and other holdouts switch to systemd, I'll probably migrate to some sort of *BSD, though I would be doing so with a 'heavy heart'.

Finally, an observation that comes from experience: in linux, just as in life, we have to guard against the tragedy of the commons. So if a systemd free fork appears (and does well), systemd dies out because of its inherently flawed nature, or those of us who care migrate to something else, let's not be complacent and allow for some individual's ego fuck things up for the rest of us.

Best,

Alex

P.S. I can think of a certain developer who should read said essay. Of course, like many people, they'd find some way to dismiss its message in this instance while embracing it in so many others.

[khayyam]

Finally, an observation that comes from experience: in linux, just as in life, we have to guard against the tragedy of the commons.

evoweiss ... the Lloyd/Hardin argument is weak, you'd have to accept the Malthusian view that events in nature are determined (cumulative, intrinsic, etc), rather than relational, and so skip the game-theoretic insight of the subject as much determining as determined. The only tragedy here is that this argument mistakes cause for effect, and it does so in a way that entirely obfuscates the issue (whether that issue be where a particular demarcation of private/public might be drawn, or what level of population, or growth, can be managed/controlled via planning) and so how "technical problems" can be understood.

best ... khay

[mrbassie]

Illumos Dev Bryan Cantrill, speaking of nomanclature, rants about sytemd. Got to 48:55.

https://www.youtube.com/watch?v=Ya6h2zKlpaQ


EDIT: comes up again at 1:23:00

Fun interview.

[gwr]

Finally, an observation that comes from experience: in linux, just as in life, we have to guard against the tragedy of the commons.

evoweiss ... the Lloyd/Hardin argument is weak, you'd have to accept the Malthusian view that events in nature are determined (cumulative, intrinsic, etc), rather than relational, and so skip the game-theoretic insight of the subject as much determining as determined. The only tragedy here is that this argument mistakes cause for effect, and it does so in a way that entirely obfuscates the issue (whether that issue be where a particular demarcation of private/public might be drawn, or what level of population, or growth, can be managed/controlled via planning) and so how "technical problems" can be understood.

best ... khay

And yet, I see a superficial similarity with the Tragedy of the Commons. You certainly do have a small group of people in a large, common code base that are solely reaping some benefit while everyone else is reaping the detriment.

Edit: grammar.

[khayyam]

And yet, I see a superficial similarity with the Tragedy of the Commons. You certainly do have a small group of people in a large, common code base that are solely reaping some benefit while everyone else is reaping the detriment.

gwr ... but again, mistaking cause for effect ... such outcomes are not due to the commons per se but due to relations between various agents using (or having access, or rights to) those resources. When people speak of "the tragedy of the commons" it is a statement based on the question of shared versus private access, with the argument that it is by virtue of them being shared that is the cause of the tragedy. The argument is that were they under the rule of private access they would be protected by self-interest ... something claimed to be lacking in a commons. I won't go into it in more detail, but why is self-interest not in operation (seen from a game-theoretic perspective) when there is no singular claim? The question you are posing above is about control, authority, right, etc, and not about why commons may fail.

best ... khay

[Tony0945]

And yet, I see a superficial similarity with the Tragedy of the Commons. You certainly do have a small group of people in a large, common code base that are solely reaping some benefit while everyone else is reaping the detriment.

gwr ... but again, mistaking cause for effect ... such outcomes are not due to the commons per se but due to relations between various agents using (or having access, or rights to) those resources. When people speak of "the tragedy of the commons" it is a statement based on the question of shared versus private access, with the argument that it is by virtue of them being shared that is the cause of the tragedy. The argument is that were they under the rule of private access they would be protected by self-interest ... something claimed to be lacking in a commons. I won't go into it in more detail, but why is self-interest not in operation (seen from a game-theoretic perspective) when there is no singular claim? The question you are posing above is about control, authority, right, etc, and not about why commons may fail.

best ... khay

I thought The Tragedy of the Commons referred to exploitation of the Commons by all and no stakeholder to replenish the Commons.

[mrbassie]

@Mods delete please. Quoted myself instead of editing.

[khayyam]

I thought The Tragedy of the Commons referred to exploitation of the Commons by all and no stakeholder to replenish the Commons.

Tony ... well, I don't really want to give the term much in the way of credence, but to take that at face value, do we not all "exploit" resources (in the purely technical meaning of the term)? Are such exploiters of the commons not stakeholders? What is it that causes them to fail to "replenish" (if indeed such a thing happens)?

best ... khay

[Shamus397]

@Tony0945: Khayyam's analysis of the so-called 'Tragedy of the Commons' is a good one. You have to remember that it was written to be used as an excuse for those who wanted the Commons for themselves only. The fact that it seems to be remembered as a cautionary tale shows that its power to propagandize is still potent to this day. It's an excellent example of 'when those in power give a reason for something, there are always two reasons--a good reason, and a real reason'.

A good thing to keep in mind with all things SystemD as well.

[evoweiss]

Khay, Shamus,

Pardon me for asking, but how are you two interpreting my post? I hope it's not that Linux should somehow be privatized; nothing could be further from the truth. I really don't see why you'd come to that conclusion.

As per the commons, yes, I know, to quote Khay, that "such outcomes are not due to the commons per se but due to relations between various agents using (or having access, or rights to) those resources." It was the latter I was referring to.

Best,

Alex

[Tony0945]

Tony ... well, I don't really want to give the term much in the way of credence, but to take that at face value, do we not all "exploit" resources (in the purely technical meaning of the term)? Are such exploiters of the commons not stakeholders? What is it that causes them to fail to "replenish" (if indeed such a thing happens)?

best ... khay

Wasn't it like everybody grazes their sheep there but nobody acres about overgrazing? I'm not sure. Need to google more. It may have different meaning now.

[Shamus397]

Wasn't it like everybody grazes their sheep there but nobody acres about overgrazing? I'm not sure. Need to google more. It may have different meaning now.

Decent analyses can be found here, here and here; in spite of where they come from. This may also be of interest.

Yes, the 'Tragedy of the Commons' is a myth, and one that is long overdue in dying.

[khayyam]

Tony ... well, I don't really want to give the term much in the way of credence, but to take that at face value, do we not all "exploit" resources (in the purely technical meaning of the term)? Are such exploiters of the commons not stakeholders? What is it that causes them to fail to "replenish" (if indeed such a thing happens)?

Wasn't it like everybody grazes their sheep there but nobody acres about overgrazing? I'm not sure. Need to google more. It may have different meaning now.

Tony ... this is the idea that its intended to invoke in the reader, yes, but again, why would they not care, surely they have a vested interest (as users of the resource) that overgrazing doesn't occur? To say "well, some will take advantage unless ownership is clearly delineated" is to overlook that right is still involved (and so claims, etc), and that each "stakeholder" has a vested interest in making sure the resource isn't over-exploited ... in fact there is more pressure for that to be the case as such behaviour must be negotiated between the involved parties, and so provides checks and balances against abuse. Yes, such "technical problems" (to use Hardin term) don't magically work out, they require (or involve) a whole slew of things that are not reducible to whether they are 'shared' or 'private' ... and that is the point wrt to this particular idea, it claims the tragedy is caused by the lack of exclusion, rather than the many other factors that the "problem" involves (social arrangements, agreements, etc).

Pardon me for asking, but how are you two interpreting my post? I hope it's not that Linux should somehow be privatized; nothing could be further from the truth. I really don't see why you'd come to that conclusion.

evoweiss ... I don't think I'd come to any conclusion in that regard, but the shared/private is exactly what the ToTC sets out to explain ... the "tragedy" being an exclusive negative for the shared component of the equation.

best ... khay

[Tony0945]

Perhaps apropos is something I remember long ago from English class in High School.

The law in its wisdom doth condemn the thief that steals a sheep from the Common.
But let's free the greater thief who steals the Common from the sheep!"

IIRC, it was about the English Enclosure Acts.

[khayyam]

The law in its wisdom doth condemn the thief that steals a sheep from the Common. But let's free the greater thief who steals the Common from the sheep!"

Tony ... variously attributed to Aesop, and others whom I can't now remember, and btw, its "goose" (to fit the rhyme) ...

The law doth punish man or woman
That steals the goose from off the common,
But lets the greater felon loose
That steals the common from the goose.

... there are many variations on this theme ... you need hope a goose is worth less than eight pence ;)

No thief be spared [death] who may be taken red handed if he is older than 12 years and has stolen more than eight pence.

best ... khay

[tld]

Illumos Dev Bryan Cantrill, speaking of nomanclature, rants about sytemd. Got to 48:55.

https://www.youtube.com/watch?v=Ya6h2zKlpaQ


EDIT: comes up again at 1:23:00

Fun interview.

Thanks for that! I'll have to find time to watch that whole thing.

I loved the rant about "libsystemd"...how if you name a library lib<my-daemon-name> you're doing it wrong and have no idea what a library is for ...so true.

Thanks.
Tom

[digi_owl]

Because they want everything to tie to systemd, so they lump all the "rope" into libsystemd.

All in all i find myself thinking "and they claim Torvalds' emails are bad?".

[steveL]

Good coders readily admit their mistakes, because they make so many of them all the time; that's what coding is about: constant revision after review.

Thanks , that one sentence says it all ! Can I borrow it for a sig?

Heh, sure, if you like. It's a summary from work, really, "just words."

I won't pretend to know the systemd end-game, but I'll suggest that their actions make more sense if you look a bit down the road.

I believe they're looking toward the day of "containerized applications."
..
I expect containerized applications to be part of a deployment strategy. Quite some time ago, the systemd folks were talking about being able to roll systems back and foward using btrfs snapshots. Corporations like to lock down the PCs they give employees. Software companies like to lock down the applications they sell. They get a big boost forward with containers and the systemd stuff underneath. They basically turn your PC into a cloud-adjunct.

It's not the PC as we know it.

Corporation may lock down PCs, but they tend to want them to be secure: which is why they're locking their network down in the first place.
As discussed, installing the systemdbust-polickysh1t attack-vector for that would be as foolish as going back to Windows in the deluded hope that it could improve data security.
Even RedHat couldn't make systemdbust work; just like Microsoft couldn't make hotmail run on Windows.

I tend to agree with your speculation, the containers bring a new light to the whole system!?%!WTF discussion although of course its still speculation at this point? They would have a lot of work to do to make this possible and stable, although RH tends to just push things out to enterprise anyway, regardless. (Im not flaming there, just see it everyday) Correct me if im wrong, actually it just sounds like an expensively coded sandboxed app that has been available to any hardened BSD or Linux for years. Any real *nix veteran in the hosting space knows this. I guess the key is the automation or replication maybe? Funny thing is its all been available for years already!
*edit meant to say expensively coded jail for BSD*

So how exactly does it "bring a new light", when it's exactly the same idea?

I agree wholeheartedly that all we're talking about is Linux finally catching up to BSD jails.

The pretext that it's "all new" "innovation" is the problem: it's nothing new, just YAF marketing gimmick to make newcomers think they are being let in on a mystery, which is the same crappy approach Poeterring always takes: obfuscation behind a flurry of sales-pitch, designed to appeal to teenage males without a social-life; who really should just get out more, afaic.

Fanbois: just remember, in a few years you'll have grown out of your need for some weirdo's approval.
Why not skip ahead to the fun stuff?

[steveL]

David Edmundson from KDE, concluded that "In many cases [systemd] allows us to throw away large amounts of code whilst at the same time providing a better user experience. Adding it [systemd] as an optional extra defeats the main benefit".

http://blog.davidedmundson.co.uk/blog/s ... and-plasma

Looks like more 1:1 integration directly with systemd without any room for other alternatives.

Yup; the sad part is how lame that post is, yet it will be used as pretext: "see, we have another KDE developer putting the case for systemdbust, it must be right.."

The reason it is so lame, is because all of the use-cases he discusses, should have been worked out at an ABI level, way back when kde-4 was being conceived, and they dropped the far-better and more performant dcop for dbus.

ABI level, in the same way that LADSPA quietly sorted out pro-audio plugins, on a cross-platform basis: specifying a useful API, simplifying to the bare minimum required, while not leaving out anything needed. Only needing one revision in a decade, to take account of usage-feedback.
LADSPA has been superseded by LV2, with extensions. The interface is still very "simple".

The truth is that not everything even needs a random desktop bus; and where it does, we should use TIPC, a proven (cross-platform) pub/sub channel that is already in the kernel (for over a decade.)
(k)Dbust developers were told this back in 2012, so there is no excuse for their shameful avoidance of prior-art.

For instance, the simplest application would be email notification; how the backend scans for, or receives, new data is irrelevant. A simple specification of the functional API needed, which any group of Gentoo users (on real hw) could do, add some supporting data structures, and leave it at that.
Any lib can provide that functional interface, and it can interface to any process providing the backend data, however it likes.

Personally I quite like the idea of an inotify watcher; and I don't see any need for a fandango of crappy network-protocol undesign shoved onto the local machine to facilitate GPL-evasion, when basic POSIX IPC is much cleaner, and works cross-platform.

Still, whatever floats your boat: just stop pretending that a complete lack of understanding of modularity and userland *nix programming is an excuse for idiotic mis-design.

If you want pub/sub, so the email watcher can just send to N random processes it does not need to know about, use TIPC already.

Though really I only want one "client": the systray applet displaying the data. And it seems simpler just to let it get its own info, using the lib and no other processes.
And doesn't that just sound like your email client?
Others may differ ofc: that's the beauty of an ecosystem. Let it work for us.

[Ant P.]

I think recompiling the kernel to add TIPC is overkill. The good kind of overkill maybe, but you're not going to sell bloatmakers on something “different” they can't understand.

Dumb it down to ideas they already know, say Avahi over loopback, and they might be a bit more enthusiastic.