|
|
Log in / Subscribe / Register

Colin Watson resigns from Debian Technical Committee

[Posted November 17, 2014 by ris]

Colin Watson announced his resignation from the Debian Technical Committee before Russ. "I appreciate that the timing is such that this looks like a response to Joey's mails, or perhaps to some other recent discussions. That isn't the case. I've been doing a good deal of refactoring of my life recently as a result of realising that I was burning out, and right now it's important that I make an effort to spend my Debian time on things I find relaxing rather than things I've been finding stressful." (Thanks to Jeff Schroeder)
to post comments

Not only Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 19:30 UTC (Mon) by brodo (subscriber, #4049) [Link]

Not only him, it seems, but Russ Allbery as well. https://lists.debian.org/debian-ctte/2014/11/msg00071.html

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 19:36 UTC (Mon) by BrucePerens (guest, #2510) [Link] (121 responses)

See the entire discussion here.

As you can see, I found out this morning that anyone who criticizes systemd has a character flaw. It's not a technical or policy argument at all. You can't really even start talking about it without being criticized for the tone of your discussion.

This is mad, folks.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 19:40 UTC (Mon) by josh (subscriber, #17465) [Link] (57 responses)

There are many high-quality critiques and critics of systemd out there, who get respect from both sides.

You are simply not one of them. And the fact that you felt the need to quickly post comments on three successive posts about people having to leave Debian or positions of responsibility in Debian over this, choosing not to acknowledge any of the reasons why, and instead to launching yet another attack about systemd and the people working on it, demonstrates one of *many* reasons why.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:12 UTC (Mon) by BrucePerens (guest, #2510) [Link] (56 responses)

I think you are expecting me to argue against systemd on a technical basis and you are disappointed that I do not. But this isn't really a technical argument. It's process, policy, and marketing.

It is not how any of the individual pieces work, but it is the fact that they are in a project as big (maybe larger) than the KDE or GNOME GUI which the rest of the world has been expected to accept en masse rather than ala carte.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:32 UTC (Mon) by josh (subscriber, #17465) [Link]

I'm expecting you to make useful, actionable criticisms about what in systemd you don't like, in ways that could actually be fixed. If you can point to some specific issue you have with systemd, that can be addressed. And in particular, anything you *think* you could address by a process change could potentially instead be addressed by fixing the code to achieve the same result.

Saying systemd needs to be developed by a battling set of adversarial teams in separate repositories, on the other hand, ranks in usefulness right around saying that it should be developed by people who park their bicycles in blue bikesheds. And it's particularly inappropriate coming from someone not actually doing work here.

There exists a set of people willing to do an enormous pile of work that nobody else has, willing to confront problems that most people just work around, and thus-far putting up with an enormous amount of snark and criticism (and worse) for doing so. And you're going to complain that they work too well together and ought to be broken up into separate projects, to satisfy some philosophical point that you care about enough to argue about but not enough to actually work on yourself?

If you don't like systemd, don't "accept" it; work on alternatives. Also bear in mind that the people working on systemd in Debian in particular have gone out of their way to keep a use case that you seem to care about alive, at significant additional effort, when other distributions have not bothered doing so. They could use some help, but you don't seem to have any help on offer.

More importantly, whether you have process criticisms or technical criticisms, or just snarky attacks, take them to some forum *other* than the comments on an article about how yet another developer has suffered attacks about systemd.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:32 UTC (Mon) by josh (subscriber, #17465) [Link] (43 responses)

Oh, and to be clear: the process, policy, and marketing is working *just fine*, thanks. If you don't like the result, perhaps you should work on your process, your policy, or your marketing.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:51 UTC (Mon) by BrucePerens (guest, #2510) [Link] (42 responses)

the process, policy, and marketing is working *just fine*, thanks

Well, I for one am not compelled. I used to be. When multiple Debian developers begged me to intervene in this, I did not. I only cared that my system booted faster. But eventually the emotional factor came to a head. Too much angst in too many places. Astonishingly many. And I could no longer ignore it.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:00 UTC (Mon) by niner (guest, #26151) [Link] (2 responses)

And instead of educating people and soothing the fears, you opted to fuel them even more. Great way to abuse a position.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:06 UTC (Mon) by BrucePerens (guest, #2510) [Link] (1 responses)

Well, I think something's wrong, and I want to motivate people to rethink. This is hardly abuse.

It's not as if I will profit from this. I've blown much of what would have been a billable day discussing this, and will be most of the evening catching up. But I have strong feelings and I am hardly the only one.

I am seeing no admission of a drop of fault of any kind on the other side.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:34 UTC (Mon) by josh (subscriber, #17465) [Link]

> I am seeing no admission of a drop of fault of any kind

Then you haven't been looking nearly hard enough.

Every day, I see constructive bug reports filed, solutions discussed, real issues fixed (sometimes in systemd, sometimes in packages using or interacting with systemd). I see people on both sides expressing regret that this issue has caused so much strife, and frustrated that they have little to no ability to stop it. I see people learning new things, listening to new use cases, and improving software to make sure things run smoothly for everyone. I see people bending over backwards to support use cases that don't personally affect them. I see people trying to quell the flames, regardless of their source. I see people writing heartfelt hopes that after all is said and done, regardless of the outcome, everyone can go on with their lives and treat each other with compassion.

And I see people showing up with more fuel to throw on the flames.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:25 UTC (Mon) by rahvin (guest, #16953) [Link] (1 responses)

You appear to be arguing in a roundabout way that the politics of the discussion trump the technical merits and that software should be designed around political issues rather than technical.

I disagree and you will find many people disagree.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 9:49 UTC (Tue) by epa (subscriber, #39769) [Link]

"Teach the controversy..."

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:26 UTC (Mon) by josh (subscriber, #17465) [Link] (36 responses)

I'm not talking about the piles of people flaming each other over systemd; that certainly needs to stop; it's taking away time that could be better spent either developing systemd *or* developing alternatives (e.g. systemd-shim, cgmanager, upstart). I'm talking about the process, policy, and marketing that has successfully demonstrated an impressive development and adoption rate for new sets of technologies, almost as though they solve problems people actually have and care about. *That's* working just fine; to some extent it has attracted a pile of detractors precisely because it has worked quite well.

If you've been asked to calm tensions and reduce the flames, you're doing an exceptionally poor job of it.

You already *know* that the systemd developers will not fracture themselves into separate adversarial projects to satisfy your whims. Even *if* they decided for legitimate reasons to split their projects across multiple repositories, you can hardly force them to stop contributing changes across multiple such repositories, nor should you even want to stop productive contribution. Suggesting that cannot possibly lead anywhere constructive.

Now, suggesting specific issues in systemd, or even a pattern of such issues, *that* might lead somewhere constructive. It might even result in such issues being fixed, if others agree that they're bugs.

If you're speaking to people who might develop alternatives, then speak to people who might develop alternatives, rather than speaking rhetorically at systemd.

And even better, if you have any suggestion to step the set of obnoxious trolls on both sides that demoralize everyone involved, we're all ears. (I certainly wouldn't claim that only one side has zealots. I would object to the claim that either side consists entirely of zealots, as well.)

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:37 UTC (Mon) by BrucePerens (guest, #2510) [Link] (35 responses)

If you've been asked to calm tensions and reduce the flames, you're doing an exceptionally poor job of it.

All of the people who contacted me wanted me to help put on the brakes.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:42 UTC (Mon) by rodgerd (guest, #58896) [Link] (27 responses)

The lurkers support you in email, eh?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 16:44 UTC (Tue) by wookey (guest, #5501) [Link] (26 responses)

"The lurkers support you in email, eh?"

I think it's pretty clear by now that a lot of people are very unconfortable with the 'systemd direction', so this is a very weak criticism.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 18:45 UTC (Tue) by rgmoore (✭ supporter ✭, #75) [Link] (1 responses)

I think it's clear that some very vocal people are very uncomfortable with systemd, but one of the important lessons of the internet is that it's hard to judge the size of a group by the volume of its comments. Noisy minorities, even really tiny ones, can dominate a discussion far out of line with their numbers just by posting all the time. And, as these Debian resignations show, when that noisy minority is sufficiently awful, it can drive people of good will out of the conversation because they're unwilling to put up with the spew from the people trying to derail the conversation.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 18:51 UTC (Tue) by dlang (guest, #313) [Link]

It's also false to assume that just because they are making noise they must be a small minority.

I fully agree that the anti-systemd people are a minority.

But I will also agree that the pro-systemd people are a minority.

the vast majority of users don't care.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 18:56 UTC (Tue) by rodgerd (guest, #58896) [Link] (23 responses)

I fail to see much real evidence of that "lot of people". And I certainly see no evidence that a "lot of people" think Bruce is the hallowed saviour of Linux in general and Debian in particular.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 19:06 UTC (Tue) by dlang (guest, #313) [Link] (8 responses)

> I fail to see much real evidence of that "lot of people".

so what would you consider 'real evidence'?

> And I certainly see no evidence that a "lot of people" think Bruce is the hallowed saviour of Linux in general and Debian in particular.

nice strawman there, but there's nobody claiming that.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 20:49 UTC (Tue) by rodgerd (guest, #58896) [Link] (7 responses)

>> I fail to see much real evidence of that "lot of people".

> so what would you consider 'real evidence'?

Well, I'd expect at a *technical* level to see significant pushback from DE projects against relying on tooling like logind. I'd expect to see many distributions choosing *not* to implement systemd. I'd expect to see a lot more people who, like you, have criticisms derived from meangingful attempts to use systemd.

Instead it's a fairly small echo chamber of the same handful of people making noise across a common set of online forums. And the people who do work that copuld be affected positively or negatively by systemd are mostly adtoping it (as programmers or distro maintainers).

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:08 UTC (Tue) by mgb (guest, #3226) [Link] (6 responses)

> Well, I'd expect at a *technical* level to see significant pushback from DE projects against relying on tooling like logind.

Gnome and KDE are having great fun with their experiments - which is their right as volunteers - and continuously shedding disgruntled users as a result.

Linux is the most widely used O/S overall. It is also the leader on servers and mobile. Gnome and KDE have gotten nowhere.

Linux successes derive from the Unix Philosophy which enables innovation. Monolithically entangled systemd is the opposite. Android ignores systemd and the pushback you're seeing against systemd is from Linux's other success story - server sysadmins.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 1:40 UTC (Wed) by ebassi (subscriber, #54855) [Link] (5 responses)

Linux successes derive from the Unix Philosophy which enables innovation. Monolithically entangled systemd is the opposite.

this is monumentally wrong on various levels already, but the cherry on top is the mention of Android. newsflash: Android is very much monolithic — and it's getting even more calcified with each release, now that Google is trying to rein it the whole ecosystem.

you clearly have absolutely no idea what are you talking about.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 2:14 UTC (Wed) by mgb (guest, #3226) [Link] (4 responses)

Android is an important application of Linux that was made possible by the Unix Philosophy. But Android itself is monolithic and therefore not the future of Linux.

Systemd too was made possible by the Unix Philosophy. But systemd itself is an entangled monolith and therefore not the future of Linux. Systemd might remain a critical component of the Gnome/KDE world but the Gnome/KDE world will then always remain an insignificant fraction of the Linux world.

If Linux Desktop ever succeeds as Linux Server and Linux Embedded and Linux Mobile have succeeded, it will be via a route that supports innovation, not via a systemd monolith that raises barriers to innovation.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 8:52 UTC (Wed) by smurf (subscriber, #17840) [Link] (3 responses)

Please explain why having everything in one git repository you can freely clone and modify and whatnot, as opposed to ten git repositories …, hampers innovation.

Please explain why that should even be a problem, given the fact that almost all of the protocols involved are completely open. So is the process of developing them; it's not as if Lennart just shakes his head vigorously and the dbus spec for timedated falls out of his left ear.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 9:07 UTC (Wed) by mgb (guest, #3226) [Link] (2 responses)

Git doesn't hamper innovation.

Systemd's monolithically entangled design hampers innovation.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 9:42 UTC (Wed) by niner (guest, #26151) [Link] (1 responses)

Repeating the same claim over and over does not explain _why_ systemd's design supposedly hampers innovation.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 17:08 UTC (Wed) by nix (subscriber, #2304) [Link]

It also doesn't explain why it hampers innovation any more than having a standard set of any other APIs, say, like POSIX, or libc in general, or libX11, or the GNOME or KDE APIs, hampers innovation. It's not as if switching out systemd's pieces is impossible: indeed, with the exception of PID 1 it's pretty easy.

The only problem, really, is that many (not all!) of those pieces seem to depend on PID 1 being systemd, and that's probably not impossible to solve: you'd just need to write another program (not necessarily PID 1) that did the same things systemd does with cgroups and DBus interfaces -- like, say, the already-existing systemd-shim. The argument must therefore be that systemd hampers innovation until someone makes a new program that mimics some of its behaviours, whereupon it suddenly would cease to hamper innovation without changing in any other sense, seems likely to be somewhat flawed. Particularly given that this program already exists.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 20:02 UTC (Tue) by notninjaz (guest, #99725) [Link] (13 responses)

I'm just one person, but back in 1999 when I was interviewed by Business Week about the rise of Linux, Bruce Perens is the one person who I reached out to when they started asking difficult questions about the ideology of Free Software vs. the practicality of incorporating non-free elements. He was the one I immediately thought of as someone could be trusted to navigate a delicate situation in a way that would lead to a positive outcome.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 20:44 UTC (Tue) by jspaleta (subscriber, #50639) [Link] (12 responses)

People change.

Can you imagine that 1999 version of said person encouraging deliberate Balkanization of any development community? Deliberate creation of smaller groups who are openly hostile to each other, as a positive thing to see?

I can't. I actually can't fathom anyone who is thinking rationally to want to see that. But you know my family is actually from the Balkans, and chose to flee to the US to avoid exactly the hostility. Maybe I just have a different perspective...as in I still have a modicum of compassion and empathy.

No matter how systemd is being managed or constructed I would never, ever, ever wish "Balkanization" on any community or project and the hostility inherent therein.And I pity the human being who has reached the point in their life where they feel creating a pervasive hostility between sub-communities with more similarities than differences is the preferred approach as a open development model. It's one thing to agree to disagree, and to be civil about making space for that disagreement, but to prefer open hostility? And to prefer to take a project with active development and to create internal hostility? That's just... terrible. To want to see more fighting and less cooperation.. less consensus? Is this the person you talked to in 1999?

Whomever existed in 1999 is not the person that is talking now in 2014.
Maybe he's recently been replaced by a Skrull. I actually hope that's the case. Because if he has been replaced by a Skrull, we might be able to get the original Bruce back if we could just find where he's being kept.

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 20:59 UTC (Tue) by notninjaz (guest, #99725) [Link] (9 responses)

jef,

After using Debian, I started using FreeBSD for non-enterprise servers. That community had a similar disagreement to this one which led to the DragonflyBSD fork. I consider that as a positive development compared to accusing people of racism over disagreements about technical direction.

When I stopped using Debian, it was because it had become clear to me that it was first and foremost a social activity for its developers. I ended up using Mandrake because it supported the video cards on the market at the time after the one in my workstation died, even though I would have preferred something like Ubuntu (ie., Debian for Use Cases)

For my part, I think good judgement is timeless and not tied to the question of the day, and I'm grateful for Bruce's involvement. It's apparent that he knew what he was getting into by offering anything but praise for systemd and the fact that he still got involved is commendable as I see it.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:13 UTC (Tue) by jspaleta (subscriber, #50639) [Link] (8 responses)

You know, there's a wide gap between arguing sincerely for more loose coupling between subprojects and being a proponent of Balkansian of a project. A wide gap.

Anyways, as a current FreeBSD user I hope you are looking forward with excitement to the continued development of systembsd and its future integration into your current operating system of choice.

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:20 UTC (Tue) by jspaleta (subscriber, #50639) [Link] (6 responses)

Sadly LWN doesn't really cover qnx or android so much so you are correct.

I'm actually currently trying to help one Deb maintainer craft a more optional systemd native init and close a Debian bug. But I'll probably just make it worse.

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:29 UTC (Tue) by johannbg (guest, #65743) [Link] (5 responses)

The bug number?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:52 UTC (Tue) by jspaleta (subscriber, #50639) [Link] (4 responses)

768168
Specifically this message is details the jumping in point that I was interested in trying to help with:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=768168#60

I haven't actually proposed anything on the bug yet..I'm having a sidebar conversation in email at the moment, doing so testing on my fedora box. If you have the time and have suggestions on how to get the necessary functionality encoded I think that would be great..

The fedora watchdog units don't cover the situation where transiton to wd-keepalive is needed as far as I can tell.

main issue: How to gracefully handling falling back and running wd-keepalive when kernels are so configured to ignore the device close that watchdog does.

The existing deb initscript starts wd-keepalive when watchdog service is stopped if the watchdog config file says keepalive should be running if watchdog isn't.

So.. under sysv init wd-keepalive gets started conditionally when watchdog is manually stopped..but also as a result early in the shutdown process as the sysvinit stop function gets called then as well. Because for such configured kernels if watchdog is stopped early in shutdown.

Wrinkle the first.. watchdog needs to be started "late enough" if configured to watch other services. if watchdog starts too early before those services it will think they are dead and do its watchdog magic inappropriately. So yeah basically a race. if all watched services have systemd units with correct Before=watchdog using local admin overrides in /etc/systemd/ as per best systemd practises, then watchdog should start late enough without any issue. But if some of those services are started by sysvinit compat and not by systemd units can a watchdog unit be prepped that runs late enough as a catchall for watched sysvinit services? And Which is to be expected for people upgrading to a systemd init in jessie I think.

Wrinkle the second...watchdog needs to be taken down early in shutdown before any watched services are dropped and wd-keepalive needs to conditional run in its place. Can we get a special early shutdown unit to do that conditionally starts up wd-keepalive and conflicts with watchdog?

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 23:46 UTC (Tue) by johannbg (guest, #65743) [Link] (3 responses)

I'm not sure where to begin actually after looking into that bug

EnvironmentFile=/etc/default/watchdog <---

Calling an environmental file to load an environmental used later in the unit is a no no and obsoleted regardless if it's called /etc/default/foo ( Debian ) or /etc/sysconfig/bar ( Fedora )

Admins overwrite package default configuration entries either via conf snippets or a copy of a full blown units in /etc/systemd/system so there is no need for that "Say YES","Say NO" or OPTIONS="" which these files usually consist of.

Let's take the next ugly line

ExecStartPre=/bin/sh -c '[ -z "${watchdog_module}" ] || [ "${watchdog_module}" = "none" ] || echo /sbin/modprobe $watchdog_module'

For the first you dont escape to shell in units, secondly you dont load modules with units either, you drop a module load snipped into /etc/modules-load.d/<foo>.conf or better yet built it into the kernel

But if you insist on ignoring what I said here above create a shells script and call that instead of escaping and if you are going to be loading a module at least be nice about it

ExecStartPre=-/sbin/modprobe -qab <foo>

And never unload a module in a unit yes that means never.

Unloading modules is a debuging/developer/testing tool and most certainly should not be done during shutdown of the unit.

Now since you are in communication with him I want you ask him this.

Are you going to be using systemd in your Debian instalment?

If the answer to that is yes then ask him if he simply cant use the built in hardware and software ( for units ) watchdog support that comes with systemd?

That built in watchdog support has been there since the ages and was implemented you know for that embedded/server crowd which some people have apparently stated on numerous occasions does not use systemd.

Yes we are aware that/dev/watchdog is single-user only and it defaults to off in systemd so an separate external watchdog daemon, ( such as watchdog ) can be used ( nobody is removing choices here ).

That takes care of his ( the reporter ) problem but not the said maintainer(s) one(s) since what we are seeing here above ( which is the guts of that hackish unit outlined in that bug report ) is a complete fallout from the gorilla warfare that has been taking place in the Debian community namely distracting it from forming solid packaging policy, separating ( on a package level ) the $other-inits functionality ( which the I want to support every init system out there has to arguably come up with ) and systemd.

Now the watchdog maintainer needs to support either two packages or broken integration with systemd in the form of hackish units containing environmental files which may or may not have to load modules and what not.

Btw this can simply be a ( race ) bug as well which we found out [1] when we migrated stuff like this in Fedora

1. http://svn.savannah.gnu.org/viewvc/trunk/bmc-watchdog/bmc...

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 0:12 UTC (Wed) by jspaleta (subscriber, #50639) [Link] (2 responses)

I'm actually trying to redo the units from scratch. I'm not trying to keep the existing units because of the problems there.

With regard to systemd's watchdog.. I think the idea is making sure...if admins migrating existing systems to jessie have made a self-determination that they want the existing watchdog to work until such time as they can transition their local config to using systemd's watchdog confidently... then the watchdog service should still work for them during that transition.

I don't think there's a clean way to automate that transition for admins. I think there's going to be a real need to help admins keep their existing watchdog service working so they can be confident in transitioning to systemd's watchdog. I think this is probably most true for local service inits that are going to still be using sysvinit until the local admin transitions those over and not inits in the debian repo and make use of the systemd servie lifetime management to take action on service crashes and whatnot.

So for me really this is about given those admins some breathing room and time to transition any local configs they need over to systemd so they learn to use systemd's native units for service management including the FailureActions directives displacing watchdog's software supervision role.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 0:52 UTC (Wed) by johannbg (guest, #65743) [Link] (1 responses)

Hmm why dont you just leave the legacy sysv initscript as is if the intent is to give the admins some "breathing room" and they just continue to use that?

When they are ready to advance themselves they simply use the built in watchdog feature.

But have at it if you think this serves some purpose...

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 0:59 UTC (Wed) by jspaleta (subscriber, #50639) [Link]

Leaving it as a sysvinit script only might be the best course of action, I was considering suggesting that actually.

But as the attempt was started to convert it over and fedora already had a converted unit file.. I was going to assume there was a need for it to work as a native unit config.

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 4:43 UTC (Wed) by notninjaz (guest, #99725) [Link]

I'm not a current FreeBSD user. Back then, I switched to OpenBSD for most things I was doing and DragonflyBSD for another. I'm currently only using RHEL for Linux except a couple of eval systems running SLES.

The point I was hoping to make was that when a group figures out it's not one group, but actually two groups, it is usually most beneficial to just separate and pursue whatever it is that inspires each of the groups rather than resorting to brutal attacks in an effort to achieve something that looks like unity.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:09 UTC (Tue) by Zack (guest, #37335) [Link] (1 responses)

You're obviously very concerned about Debian, and specifically and repeatedly the effects of systemd adoption in Debian, and with everybody involved in it. There's also a lot of concern about Ubuntu's doings, in general.

So much concern in posts about distributions you don't actually use. Maybe that's just in your nature: to be a very concerning person.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 21:21 UTC (Tue) by notninjaz (guest, #99725) [Link]

Zack,

I'm actually not overly concerned about what happens with Debian, since it's a distribution that I don't use and don't plan on using in the future. i think open discussion about individual distributions and their goals can only benefit the Linux community in general, though.

If anything, Debian's adoption of systemd would benefit me, as Debian users tend to actually use the current release while enterprise Linux customers usually want a year or few before doing anything important with a new major release.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:51 UTC (Mon) by ovitters (guest, #27950) [Link] (5 responses)

> All of the people who contacted me wanted me to help put on the brakes.

Ah, you think you're entitled to behave like this. You seem to have forgotten that maybe your viewpoint is off. People aren't out to get you. Yet with your post you just are putting loads of fuel on something what wasn't intended to be a fire.

Your need to behave like this is just weird. Even criticize people who don't like how you're behaving. You're being very aggressive across 3 posts. Doesn't it seem like a tad strange?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 8:17 UTC (Tue) by oldtomas (guest, #72579) [Link] (4 responses)

> Ah, you think you're entitled to behave like this.

To behave like what, pray? To say that the way Systemd has been introduced *is* the real problem? Well, it *is*.

Bruce's right, imho.

> You're being very aggressive across 3 posts.

Kettle and all that.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 11:51 UTC (Tue) by ovitters (guest, #27950) [Link] (3 responses)

> Bruce's right, imho.

So one person left because of all the hate. All of that is the fault of systemd project. No responsibility lies elsewhere. Further, because some people asked him, Bruce goes out in full force. Simplistic and inappropriate.

> Kettle and all that.

I'm not trying to pretend to approach this in a nice way. Though IMO, I am.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 12:13 UTC (Tue) by oldtomas (guest, #72579) [Link] (2 responses)

> So one person left because of all the hate.

Don't get me wrong: I see unacceptable behaviour on both sides. But I don't see unacceptable behaviour on Bruce's part. Criticising systemd from a technical, architectural or (as is the case here) social perspective is not hate.

> All of that is the fault of systemd project. No responsibility lies elsewhere.

I think we can agree that such a statement can't be taken seriously. I don't think Bruce has stated that.

Why this hyperbole, then? Was it necessary? Was it useful? To what end?

> Further, because some people asked him, Bruce goes out in full force. Simplistic and inappropriate.

Bruce just pointed out what he sees as mistake the systemd project has made. In an -- as I see it -- articulate, polite way. One may agree or disagree with him. I don't see "full force", "simplistic" or "inappropriate".

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 12:59 UTC (Tue) by niner (guest, #26151) [Link] (1 responses)

So it is articulate and polite to post rants about systemd to completely unrelated news posts? To re-post the same accusations, even if they already have been refuted somewhere else?

So in your opinion, what has someone to do to flame?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 17:30 UTC (Tue) by oldtomas (guest, #72579) [Link]

> So it is articulate and polite to post rants about systemd

whatever your definition of "rant" is, it doesn't seem to coincide with mine.

> completely unrelated news posts?

Sadly, it's not completely unrelated. The deterioration of climate brought about by this conflict (and I think we agree that there are enough mistakes from both camps) is definitely related to the resignations we see. It may not be the only cause, but it sure has had its part in it.

> To re-post the same accusations

I don't see any accusations in Bruce's post. Just an analysis of things which might have contributed to the current situation, on the side of systemd, the project. You may agree, you may disagree. But accusations?

> even if they already have been refuted somewhere else?

No accusations to refute.

> So in your opinion, what has someone to do to flame?

Just go through this thread. You might find a couple of fine examples.

- "Ah, you think you're entitled to behave like this. You seem to have forgotten that maybe your viewpoint is off. People aren't out to get you."
<http://lwn.net/Articles/621265/>

- "Then it means you're hopelessly stuck in the past. Here's your social security check, grampa. Go and sit on your lawn." (luckily someone complained)
<http://lwn.net/Articles/621197/>

- "And this is how he behaves.."
<http://lwn.net/Articles/621128/>

Care for more?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 1:41 UTC (Tue) by bronson (subscriber, #4806) [Link]

Holy cow Bruce, if all you want to do is apply brakes then why show up suddenly and chuck armloads of dynamite on the tracks?

Well, it looks like the train is stopped and in a smoking heap. Hope you're proud. What a disgusting spectacle.

Any plans on getting it moving, or are you just going to disappear again?

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:41 UTC (Mon) by CopperWing (guest, #82856) [Link] (7 responses)

Systemd: 281,256 LOC [1]
Gnome: 8,811,337 LOC [2]
KDE: 22,739,295 LOC [3]

Please refrain from spreading false (by several orders of magnitude) data.

[1] https://www.openhub.net/p/systemd
[2] https://www.openhub.net/p/gnome
[3] https://www.openhub.net/p/kde

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:47 UTC (Mon) by BrucePerens (guest, #2510) [Link] (6 responses)

LOC is not the only metric. We have this difficult to quantify "importance" factor. Retraining necessary for system admins. Numbers of critical APIs. And even things like number of developers.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:04 UTC (Mon) by CopperWing (guest, #82856) [Link] (4 responses)

Could you kindly return me the favor and link here the market researches, retraining cost studies or any objective technical paper on which you are basing your evaluations?

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:20 UTC (Mon) by BrucePerens (guest, #2510) [Link] (3 responses)

I only have my own estimates at this point. Debugging what went wrong with the SystemD/GDM3/X interaction in Debian Jessie was the basis of my training estimate. It seems far from trivial to me. For marketing, I have a truism: users don't care about what they don't see.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:40 UTC (Mon) by niner (guest, #26151) [Link]

Just to be clear: your estimation is based on your personal problems with a beta version of a distribution that just recently introduced systemd? Wow, that's clearly a completely unbiased sample.

Maybe you should ask some system administrators using openSUSE. People who may not even have noticed the switch to systemd, except that commands like rcapache2 restart started to print "redirecting to systemctl restart apache2.service".

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:19 UTC (Mon) by CopperWing (guest, #82856) [Link] (1 responses)

I feel your evaluation is a little bit off, because you are focusing on Debian only.

My distribution of choice, openSUSE, switched to systemd a couple of years ago and the transition from SysVInit was smooth and almost painless.

What you are experiencing here are not systemd issues but a breakdown of the Debian's bureaucratic process, in which you have people embracing new technology and willing to do the work to support it, an old technology nobody is willing to support anymore and few specific persons into the Debian's technical Committee (which is just a consulting organ) who are abusing policy to force work on volunteers.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 23:57 UTC (Mon) by eean (subscriber, #50420) [Link]

Yea I still keep a openSUSE 11.4 installed on an old netbook, because it's pre-pulseAudio and the most dependable way to use crappy Java video conferencing software. PulseAudio faced a lot of problems since there are so many moving parts in Linux multimedia. I mean now it's overall fine (outside of a aforementioned legacy software) but it was a bumpy ride.

However my experience with openSUSE's systemd upgrade is similar to yours and not at all like ALSA->PulseAudio. I did a live upgrade to the first openSUSE with systemd and didn't even notice. I think systemd just does a fine job with legacy init scripts and so forth. I mean maybe openSUSE devs are just awesome and made it look easy, but certainly it was easy as a user.

Mostly I think process management is a walk in a park compared to Linux multimedia. ;)

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:26 UTC (Mon) by rahvin (guest, #16953) [Link]

As has already been said, it's just an init system. Most users will never even interact with it.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 23:04 UTC (Mon) by seyman (subscriber, #1172) [Link]

> It's process, policy, and marketing.

The solution is the same as for a technical argument. Fork the code and create another project running it the way you feel it should be run. If your approach is superior to systemd's, it should replace it in no time.

systemd

Posted Nov 18, 2014 9:55 UTC (Tue) by ber (subscriber, #2142) [Link]

Let us assume for a moment that the current systemd situation with Debian really is a process, policy, and marketing issue. What could we do about it?

Me for myself is missing a nice overview about all issues. Ideally it explains the pro and con arguments. Policy issues should be explained just as technical issues. An overview can help to base the discussion and the highlight the common goals.

So far I haven't found this overview and many statements in the discussion here, may they be technical or not, are hard for me to verify.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 15:11 UTC (Tue) by drag (guest, #31333) [Link]

> But this isn't really a technical argument. It's process, policy, and marketing.

Which is a utter fail on all counts.

> which the rest of the world has been expected to accept en masse rather than ala carte.

The reason people are resigning it's because of differences in opinion about systemd.. it doesn't seem so. That they simply want stop participating in Debian politics and move on with life.

Sure Debian politics is intended to fulfill a important role, but it seems that in minds of those that are resigning it is a failure. It's too much of a drain on their personal productivity and mental energy with not enough benefit. It has become a drag so they left.

If you want to abandon all discussion of technical merits in a technical decision about a operating system and focus on 'process, policy, and marketing' then that's all fine and dandy in my personal opinion, but confusing people resigning with a personal attack on 'systemd haters' just shows a lack of perceptive about what is really going on here.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:08 UTC (Mon) by lambda (subscriber, #40735) [Link] (5 responses)

You can, in fact, discuss systemd without being criticized for the tone of your discussion; I have seen plenty of bug reports filed that were responded to amicably and come to reasonable resolutions, I've seen plenty of work from people like Steve Langasek criticizing particular decisions from people doing systemd packaging and policy work, which were discussed and responded to reasonably without attacks on either side.

However, using a thread about a maintainer throwing his hands up in the air and stepping down due to countless attacks, innuendos, and and paranoia to mount another attack on the very idea of systemd is a little bit tone deaf, don't you think? A maintainer resigned from his position because of the excessively strong rhetoric used against him and the project he was maintaining, and you say that it's time to step up the rhetoric? And then posting further comments repeating the same sentiments in two further threads about resignations?

You have done some great work on Debian, BusyBox, and other tools in the past. However, you don't appear to be doing active maintenance on any of them; you've handed the reigns over to others. It now appears that you are trying to drive those others away by attacking a substantial amount of work they have done, imposing additional burdens on them, but I don't see you stepping up to do that work.

In the earlier thread, you said:

We need to break systemd up into separate projects, resolve the issues that cause the largest objections to it, and make sure it plays with others (again, including other init systems). Oh, and pushing back on the team that pushed it upon us is unfortunately necessary.

When you say "we", who exactly do you mean? Are you volunteering to fork systemd, or work with the maintainers of it to achieve this goal? Or maybe are you suggesting that Debian switch again, from systemd to uselessd or some other fork or alternate init system? When you say that "we" need to resolve the issues that cause the largest objections, which issues are you describing, that aren't being effectively managed by the Debian systemd maintainers or systemd upstream?

While you may consider a difference in preference about whether to maintain several interrelated tools in one repository vs. several a problem, that's really more a matter of preference in how a software project is run rather than a technical issue, and is only particularly relevant to those contributing to that code. If there are actual technical issues that you feel aren't being adequately addressed by upstream or the Debian maintainers, rather than just your personal development methodology and software architecture preferences, I'd be interested to hear them.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:22 UTC (Mon) by BrucePerens (guest, #2510) [Link] (4 responses)

"we" means anyone who wishes to use it while not ceding the entirety of their architectural decision set to the systemd developers.

I can not make this discussion a bug report. It is not a software bug. It is a process, policy, and marketing bug. It doesn't get fixed with a patch.

What I really want to do is send you back to sell the world a great init system, period. And then when that is done, to sell the world a great resolver (or whatever). Just not all at once.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:08 UTC (Mon) by mathstuf (subscriber, #69389) [Link]

> What I really want to do is send you back to sell the world a great init system, period. And then when that is done, to sell the world a great resolver (or whatever). Just not all at once.

Debian started this discussion a couple of years too late; this is how it happened everywhere else. Plus, it's not as if running systemd as PID 1 stops Debian from not also making the other tools which come with systemd *also* the default.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:41 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

> "we" means anyone who wishes to use it while not ceding the entirety of their architectural decision set to the systemd developers.
Nobody forces Debian to use the whole systemd infrastructure with its multiple projects.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 23:57 UTC (Mon) by judas_iscariote (guest, #47386) [Link]

No Bruce you do not get to dictate at what pace things are done or the way they are done.

""we" means anyone who wishes to use it while not ceding the entirety of their architectural decision set to the systemd developers."

Get busy then.. because the only people that can influence the architectural decisions are the people doing the work (people that include debian developers who have commit access to the systemd repository) In contrast you are just whining and lying through your teeth. that won't get you anywhere, nobody will listen to you.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 6:41 UTC (Tue) by torquay (guest, #92428) [Link]

    "we" means anyone who wishes to use it while not ceding the entirety of their architectural decision set to the systemd developers.

This is not a matter of "us vs them". There is no "ceding" of control here -- you're welcome to join the group of systemd developers, by contributing code and suggesting directions/policies on the systemd mailing lists.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:11 UTC (Mon) by smurf (subscriber, #17840) [Link] (54 responses)

Of course it's not a technical or policy argument. Your desire to split up the systemd git repo has no technical justification, and a project can declare its own policy as it sees fit; either demonstrate factual harm of that policy or kindly admit you're wrong.

Let's face it, most people attack systemd for reasons of personal preference (Monolithic! Not the UNIX way! Binary logging!) without even remotely acknowledging that there may be valid reasons to do it that way

IMHO this just reflects the fact that the pain they've gone through, learning to deal with sys5init's shortcomings during the last 20 years or so, is now all in vain and their experience (and the (not necessarily monetary) cost they incurred acquiring it) suddenly counts for nothing. So they whine and shout, instead of rejoicing that the pain finally goes away. :-P

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:26 UTC (Mon) by BrucePerens (guest, #2510) [Link] (1 responses)

I'm not here to defend SysVInit.

I am not particularly interested in opposing systemd the init system, either.

I am however very disquieted by systemd the entire base system that does very many different things, which I am to accept all at once.

Make your project policies as you wish, just don't expect the rest of the world to accept them. You are suffering so much angst from the people around you because of your policies.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 6:44 UTC (Tue) by torquay (guest, #92428) [Link]

    I am however very disquieted by systemd the entire base system that does very many different things, which I am to accept all at once.
Umm, no. The code for systemd related projects is modular. You're free to pick and choose the parts you like.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:42 UTC (Mon) by BrucePerens (guest, #2510) [Link] (51 responses)

IMHO this just reflects the fact that the pain they've gone through, learning to deal with sys5init's shortcomings during the last 20 years or so

Oh, the pain! :-) People young enough not to have watched the original Lost in Space won't get this.

Uh, the system took a few more seconds to boot. We had the dependency issue handled, not a problem. We had a bunch of more or less one-of-a-kind shell scripts where we didn't need them.

Users never cared about this issue. Developers did, either from a speed or architecture perspective.

Pain? Maybe self-imposed.

As a community, we are absolutely horrible about recognizing what users care about and what we do not. And thus we tend to work on systems programming, because that is what we know. We think that if we finally fix the system, then the users will come!

Nope.

So, I think this project has been carried out with low-empathy for outsiders from two perspectives: first there is the fact that users just don't care, and yet the project uses user needs for justification. Second, there was not sufficient thought given to how introduction of new implementations would be timed. All at once wasn't the right way to do it.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 20:59 UTC (Mon) by niner (guest, #26151) [Link] (42 responses)

"Users never cared about this issue."

Just who do you think you are that you pretend to speak for users?

I _am_ a user and I very much care about getting rid of the pile of crap that was SysV init, because I have lost just too many hours working around its deficiencies.

And I don't need someone to speak for me, thank you very much.

And stop making claims about systemd happening all at once. Just because you ignored the whole process doesn't mean that the process didn't happen. And repeating it over and over won't change that.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:11 UTC (Mon) by BrucePerens (guest, #2510) [Link] (41 responses)

I am going to give you an exercise. Stand around quietly for one hour in an Apple store, watching what goes on around you. Those are users. We're sysadmins. We are very much unlike the common people.

Look at their emotions. At what they are interested in. At the feeling of empowerment they get from that Apple stuff. At the way they absolutely lust for it. How did that happen? Why are we so far from what they are looking for?

Some of us will arrive at an understanding from doing this sufficiently. Some never will.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:15 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (38 responses)

Part of how it happened was that Apple shipped an init daemon that meant users wouldn't end up with mysteriously broken systems because of a race condition at startup.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:27 UTC (Mon) by BrucePerens (guest, #2510) [Link] (37 responses)

Matt,

In all my years working on this I have never met an init race condition on a Linux system.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:41 UTC (Mon) by niner (guest, #26151) [Link]

And because it never happened to you, it couldn't happen to anyone else, surely. But maybe that's just because you were only using distributions that did not even try to parallelize startup.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:53 UTC (Mon) by tao (subscriber, #17563) [Link]

In all my years using ext4 I have never had any fs corruption. Obviously we don't need fsck.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:56 UTC (Mon) by mjg59 (subscriber, #23239) [Link]

But you admit that they exist, right? You understand why their existence causes problems for users?

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:57 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link] (22 responses)

Then it means you're hopelessly stuck in the past. Here's your social security check, grampa. Go and sit on your lawn.

Want me to be more insulting? I can. Because I hit all those nice race conditions all the time while starting huge clusters on Amazon EC2 and other clouds.

Want details? Easy enough - the recent one involved PostgreSQL showing as 'running' while it's actually still being initialized. It was caused by unusually slow system disk and hard to reproduce otherwise. Systemd would have handled that just fine with socket activation.

Apple won because it actually cares about being reliable and easy to use. They ditched the X11 crap first (and now Linux is doing the same), they ditched the initscript mess and they wrote their own high-level networking stack (fast DHCP client, NTP client, non-braindead DNS resolver, good WiFi configuration utility, etc).

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:15 UTC (Mon) by Zack (guest, #37335) [Link] (8 responses)

> Then it means you're hopelessly stuck in the past.

Or maybe he's just competent. At least competent enough to set things up properly instead of just twiddling thumbs and waiting for someone else to provide a solution, *any* solution, and then cheerlead that blindly to mask the incompetence causing you to have to get up at night and go to the data center.

> Apple won because it actually cares about being reliable and easy to use.

What did apple win? Huge clusters in the clouds? Do they come with little cherubs?

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:17 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (1 responses)

> At least competent enough to set things up properly

sysvinit requires a degree of competence that most people (including distribution maintainers) demonstrably do not have, otherwise we wouldn't face these problems. This makes it a bad solution.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 17:10 UTC (Tue) by wookey (guest, #5501) [Link]

sysvinit requires a degree of competence that most people (including distribution maintainers) demonstrably do not have, otherwise we wouldn't face these problems. This makes it a bad solution.
It is indeed crappy in many ways. So it's quite an achievement of the systemd developers to have made a solution that many fairly reasonable people view as the worse of two evils.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:54 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

> What did apple win? Huge clusters in the clouds? Do they come with little cherubs?
Have you recently been to a tech conference in the US?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 14:29 UTC (Tue) by nye (guest, #51576) [Link] (4 responses)

>Or maybe he's just competent. At least competent enough to set things up properly instead of just twiddling thumbs and waiting for someone else to provide a solution

Wow, just wow. So anyone not building their entire init system from bone knives and bearskins flawlessly and without bugs on every system is incompetent now. Good to know.

Good job reinforcing the notion that Linux users are all elitist assholes. Great job.

>What did apple win?

A market capitalization of six hundred billion dollars?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 15:40 UTC (Tue) by Zack (guest, #37335) [Link] (3 responses)

> So anyone not building their entire init system [..] is incompetent now.

No. I'm saying that anyone who manages 30.000 servers and complains about having to get up at night because of sysvinit related problems should probably do some soul-searching before claiming other people are "stuck in the past."

> Good job reinforcing the notion that Linux users are all elitist assholes.

You mean GNU/Linux users.

> A market capitalization of six hundred billion dollars?

Which makes their operating systems more reliable? What does Apple have to do with this in the first place?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 18:34 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link] (2 responses)

> No. I'm saying that anyone who manages 30.000 servers and complains about having to get up at night because of sysvinit related problems should probably do some soul-searching before claiming other people are "stuck in the past."
I'm sure that the real Old School sysadmins read printouts of all their initscripts with their morning coffee. I freely admit that I kinda hope that simple actions like rebooting a computer after a minor update shouldn't result in completely unresponsive systems.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 17:03 UTC (Wed) by nix (subscriber, #2304) [Link] (1 responses)

Of course, just because hitting the power switch is simple doesn't mean that what it does in response is simple. Heck, what it does before running the OS is so terrifying that I'm amazed it ever gets anywhere. <http://resources.infosecinstitute.com/system-address-map-...> has a 'short' (read: way too long) primer.

Similarly, what happens after the OS boots has got more and more complex. Hence the need for things like systemd in the first place.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 19:12 UTC (Wed) by zlynx (guest, #2285) [Link]

These bootup issues are a big reason that (in my opinion) embedded ARM systems are so refreshing. They have none of that baggage. The boot loader and operating system are customized for the hardware. Flip the reset and the CPU starts executing YOUR code instead of flipping though three different ancient compatibility modes first.

And some people think we need to add ACPI to ARM? I think they're crazy.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:32 UTC (Mon) by BrucePerens (guest, #2510) [Link] (11 responses)

Apple did not ditch X11. They derived from the NeXT workstation, which used NeXT's own form of Display Postcript.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:45 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link] (5 responses)

No, it's not. Mac OS X's graphical layer was written pretty much from scratch. In fact, it's pretty much similar to what Wayland does - OS X has kernel-level modesetting and buffer management, with the compositor running as the privileged kernel client. It then composes individual window surfaces (perhaps with nice visual effects) and shows them on the screen[s].

Individual windows, in turn, can use whatever form of rendering for the GUI widgets they desire, all of it is done on the client-side. The early versions of Mac OS X in fact even had THREE widget rendering implementations: Carbon, Cocoa and Java OS X look&feel.

Display PostScript used pure server-side rendering, but allowed clients to submit short PostScript programs to render stuff. It's not even close.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 0:51 UTC (Tue) by BrucePerens (guest, #2510) [Link]

My point is that X was already gone long before Apple got this GUI. If there was X there, they put it back.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 12:26 UTC (Tue) by mjthayer (guest, #39183) [Link] (3 responses)

> No, it's not. Mac OS X's graphical layer was written pretty much from scratch. In fact, it's pretty much similar to what Wayland does - OS X has kernel-level modesetting and buffer management, with the compositor running as the privileged kernel client. It then composes individual window surfaces (perhaps with nice visual effects) and shows them on the screen[s].

And in the meantime, while everyone is looking at Wayland, X.Org seems to be quietly learning to do those things too (or KeithP is teaching it), and fixing many of its seemingly unfixable issues.

I actually originally wrote "X11" in the last sentence, but a large part of X11 seems to have become a set of shader programmes inside of X.Org and Glamor.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 18:11 UTC (Tue) by ebassi (subscriber, #54855) [Link] (2 responses)

And in the meantime, while everyone is looking at Wayland, X.Org seems to be quietly learning to do those things too

except that, no: it does not help you, unless you use a toolkit that uses the Present extension, which automatically excludes everything not bleeding edge (so: GTK+ 3 and Qt 5, if both ever get to it, as opposed to just ignoring this stuff and moving to Wayland wholesale).

in the meantime, though, you still have applications like LibreOffice, Chrome, and Firefox with their own toolkit; Eclipse, using the Java wrapper around whatever happens to run on your platform and poking Xlibs directly; or Emacs, which seems to be using GTK on the surface, except it likes to poke behind the scene, assuming it knows how X and the toolkit interact, without actually caring whether or not those assumptions hold. then we have those Motif apps from 1994…

sure: X is resilient enough to allow new extensions to land on top of that 1980s protocol, but what screws you over is the ton of legacy crap that assumes that X works in a certain way, and it won't be updated ever again, and it makes it impossible to change toolkits without breaking API or changing the soname — and then we get those hilarious posts from Linux kernel developers about how they don't break userspace ever and userspace devs should do the same if only we were competent.

if I have to spend time updating GTK+ to something modern, I'd probably spend it improving Wayland, instead of supporting yet another band aid on top of X. others may feel differntly motivated. what does not help is thinking "we can fix X11 with just another extension", because that was we thought about COMPOSITE as well, and look at where that misplaced trust led us.

X.Org and Wayland

Posted Nov 19, 2014 8:23 UTC (Wed) by mjthayer (guest, #39183) [Link] (1 responses)

> if I have to spend time updating GTK+ to something modern, I'd probably spend it improving Wayland, instead of supporting yet another band aid on top of X. others may feel differntly motivated. what does not help is thinking "we can fix X11 with just another extension", because that was we thought about COMPOSITE as well, and look at where that misplaced trust led us.

Don't get me wrong here: I am not advocating either way. On the one hand I am just observing that Wayland still seems to be in a process of maturing to cope with a lot of real life situations, and wondering whether it will still be quite as simple and clean once it has come to maturity. And on the other I am wondering how many of X11's issues will have been solved well enough in creative ways by that time, and whether at that point Wayland will still have enough impetus to replace the "just good enough" X server, which will likely by that time act much more like Wayland with Xwayland on top anyway.

Not saying that my wondering will make it so either of course, but curious to see.

X.Org and Wayland

Posted Nov 19, 2014 18:02 UTC (Wed) by Wol (subscriber, #4433) [Link]

Note that the X11 devs (Keith Packard in particular) view Wayland as a sort of X13 (X12 already exists).

The reason the X -> Wayland transition is proceeding smoothly is that there are no longer any X developers left - they've all moved to Wayland BUT. They take their responsibilities as X maintainers seriously.

Which is why "Wayland on X" is an important development focus, and they see "X on Wayland" as a very important transition mode. And why X and Wayland play nicely together ...

Cheers,
Wol

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:45 UTC (Mon) by rodgerd (guest, #58896) [Link] (4 responses)

> Apple did not ditch X11.

What colour's the sky in your world, Bruce?

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 0:54 UTC (Tue) by BrucePerens (guest, #2510) [Link] (3 responses)

The GUI on Macintosh today is derived from one that I had on my desk before some of you were born. Steve Jobs sat across the hall from me at that time and did not work at Apple. It very definitely did not incorporate X.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 1:01 UTC (Tue) by jspaleta (subscriber, #50639) [Link]

Awesome! That "before you were born" just scored me a BINGO on my LWN Bingo card I've been playing this week. I've just won a doughnut!
Awesome! Quickest BINGO ever I've gotten playing weekly LWN Bingo.

-jef

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 1:12 UTC (Tue) by rodgerd (guest, #58896) [Link] (1 responses)

Apple did, however, ditch X from the traditional Unix stack, Grandpa Simpson.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 22:29 UTC (Tue) by paulj (subscriber, #341) [Link]

No they didn't.

You've completely misparsed Bruce's comments, AFAICT. Apple never ditched X from the Unix stack they bought, as it was never there for them to ditch to begin with.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 16:42 UTC (Tue) by tjc (guest, #137) [Link]

> Then it means you're hopelessly stuck in the past. Here's your social security check, grampa. Go and sit on your lawn.

Please stop turning this into Phoronix.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 2:19 UTC (Tue) by misc (subscriber, #73730) [Link] (10 responses)

At the risk of sounding like a broken record, take a look at the bind initscript on current stable of Debian. Since it use a synchronous communication to stop bind, and then wait until the PID disappear, the PID can be reused due to a race condition ( ie, you send the communication first, but you have no way to see the pid got changed between the call to kill -0 and the rdnc call.

Saying you didn't encountered race to there is none is like saying that a bug doesn't exist because it didn't hit you.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 5:02 UTC (Tue) by BrucePerens (guest, #2510) [Link] (9 responses)

OK, I believe that race condition existed. However, this one example has been mentioned by several people. So, we had 1 bad script? In addition I understand it's better if you can use the wait system call to see if the process is really dead. So yes, that's a better fix than shell scripts. But I still wish someone could have gotten this fix in without rewriting the whole world.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 5:36 UTC (Tue) by rahulsundaram (subscriber, #21946) [Link]

The fix would have applied to one bug in one shell script. This is hardly the only one that people ever run into. It is just an easily citable example to people such as yourself who bother denying that such bugs even existed at all. Now you are left with wishing someone else fixed that problem systematically instead of using one that already did that. This is just one of the many issues with sysvinit and hence it is a deadend. Using anything else is an improvement, really. If you really dislike systemd, try openrc or something.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 5:39 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link] (5 responses)

Well, since we're on the topic of BIND initscripts... The one in Squeeze can hang indefinitely: anonscm.debian.org/cgit/users/lamont/bind9.git/tree/debian/bind9.init?h=squeeze - see line 92.

And yes, it would have been nice if reliable process tracking appeared in Linux so late and most other Unixes are still in denial.

>In addition I understand it's better if you can use the wait system call to see if the process is really dead.
Nope. Only parent can wait on PIDs and the parent is lost once the "/etc/init.d/named start" invocation ends.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 11:37 UTC (Tue) by smurf (subscriber, #17840) [Link] (1 responses)

No longer true. You can set a prctl flag that declares "I am an init-like process and get to inherit all my grandchildren".

… a flag which, not-so-coincidentally, was introduced to the Linux kernel two+ years ago – by some guy named Poettering or whatever. :-P

That by itself doesn't help, of course, if you don't know which task your process belonged to; that's cgroups' job, AFAIK first being used for this purpose by some guy named … you get my drift.

Yes, systemd has a couple of problems. Minor ones, in comparison. But somebody finally had the skill and opportunity to pull everything that's necessary for a working solution to the whole problem space into one place, instead of all the previous piecemeal attempts at it[*], and now one of the main criticisms is that it _is_ all in one place?

Sorry, but I get the distinct impression that some of these guys are just jealous that Lennart did what they couldn't.

[*] with the possible exception of upstart. However, real-life experience has demonstrated that using an event- instead of a dependency-based model doesn't work.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 17:47 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]

You can still can wait (using SIGCHLD) only on your grandchildren, not cousins or siblings. In the classic init case, you'll have to do it in the init process.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 4:12 UTC (Wed) by BrucePerens (guest, #2510) [Link] (2 responses)

Init inherits children of exited processes. But I get your point that starting it with a shell script that then exits isn't a good way to catch the pid, and anything that recovered the pid from a lock file or something at that point would be grotesque.

I accept that doing the same job of starting a daemon from 100 subtly different shell scritps is a bad idea.

But this would have taken a smaller fix than what we are currently presented with.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 4:18 UTC (Wed) by rahulsundaram (subscriber, #21946) [Link]

>But this would have taken a smaller fix than what we are currently presented with.

Some claim that but so far noone has even bothered to show the fix they have in mind. If it is so simple, why not show it?

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 9:49 UTC (Wed) by vonbrand (guest, #4458) [Link]

Many "smaller fixes" where tried over the years, none stuck. Systemd is the first one in this area to get real traction (as in across most distributions, even being embraced by the whole range from embedded though server and workstations). One has to wonder why... the Linux crowd being a individualistic, stubborn bunch of people, not so easily swayed by random marketing nonsense, should make you wonder why, and look further than conspiracy theories.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 5:58 UTC (Tue) by mjg59 (subscriber, #23239) [Link]

I wish for a lot of things, but once it's clear that the thing I think would be the perfect situation isn't going to happen I get on with dealing with what I have available to me. Nobody's ever demonstrated that this problem can be generally solved with shell scripts. The first few attempts at meaningful process tracking failed[1]. Lennart wrote something that worked. Could a bunch of the politics have gone better? Sure. But hey, if wishes were horses we'd have an enforceable trademark on the term Open Source and a fortune at the Kentucky Derby. Look at your own failures before you spend quite so much time trying to prove to the world that everyone else is fucking up.

[1] Currently dealing with a case where upstart isn't tracking all the children of a daemon because the daemon is pathological, resulting in a cascade of failures and a "killall -9" in pre-start because THIS IS INCREDIBLY DIFFICULT TO GET RIGHT

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 13:30 UTC (Tue) by cortana (subscriber, #24596) [Link]

You can only wait(2) on your children, and the running named is not a child of the process that stopping bind.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 15:06 UTC (Tue) by Doogie (guest, #59626) [Link] (1 responses)

At the way they absolutely lust for it. How did that happen? Why are we so far from what they are looking for?
Marketing. In case you missed it, Apple spent the 80's, 90's and 2000's advertising heavily the idea if you use their products you are actually a better person than people who use other products. Remember the 1984 ad? Remember the "think different" ads? Remember the "I'm a PC" ad? They appeal to people's vain desire to conspicuously own and use expensive gadgets and it works big time. Why else do you think there is a large glowing Apple logo in a location where other people (besides the owner) can see it on all their products?

Linux will probably never have that kind of marketing muscle behind it, but through open source we can achieve the same kind of software quality that Apple has been able to afford to simply pay out of pocket for. Hell, I'd say open source means we can actually do a lot better, if only we old timers and cranks would stand aside and let the passionate modern developers who are genuinely interested in actually making Linux competitive on the desktop get on with the task of doing so without having to constantly fight these ridiculous political battles.

Anyone who criticizes systemd has a character flaw

Posted Nov 19, 2014 4:19 UTC (Wed) by BrucePerens (guest, #2510) [Link]

There are two kinds of "marketing". Strategic marketing and marketing communications. Marketing communications is all of that advertising. And yes, Apple has done a lot. But the failure of the Linux desktop so far is not a matter of its not being advertised (after all, we have "viral" phenomena that work as advertising without the money these days). It's a matter of strategic marketing, and ultimately that comes down to understanding the user and what will motivate the user.

And no, this doesn't mean we have to discard those inconvenient principles, even if they cause annoying discussions. Having them is indeed our only hope of differentiation from those other products.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:14 UTC (Mon) by mjg59 (subscriber, #23239) [Link] (2 responses)

> We had the dependency issue handled, not a problem.

No. No, we didn't. Not even slightly. The fact that almost every single init system developed after sysv has had a significantly more complicated dependency model is a pretty solid indication that this issue wasn't handled. And from each iteration of that, we've learned more about how complicated this problem space actually is - see https://bugs.launchpad.net/ubuntu/+source/mountall/+bug/4... for an example of how Upstart's dependency mechanism has a corner case that can result in entirely unexpected outcomes.

> Users never cared about this issue.

Sure they did. Machines that wedged during reboots because shutting down bind correctly under all circumstances is almost impossible. Machines that forced fsck on every boot because enforcing the ordering of timezone configuration and filesystem mounting becomes difficult once you're using LVM-on-dmraid. Daemons that were still running insecure code because restarts were unreliable. In theory you could solve all these problems under sysvinit by writing sufficiently complicated code, but in practice people didn't. Making it difficult for developers to ship broken initscripts is a user-visible change that people care about.

But even beyond that - straightforward container management that's tied into your existing resource tracking system? That's a user-visible change that users care about. The ability to integrate kernel and application logging without losing early boot messages? That's a user-visible change that users care about. A startling number of systemd's features provide direct benefit to users. Arguing otherwise is just startlingly ignorant.

Is systemd some magical piece of software that's going make Linux a compelling desktop? No, clearly not. Nobody's ever argued that. But it's a necessary component for building many pieces of functionality that *are* required for a compelling modern desktop.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:33 UTC (Mon) by BrucePerens (guest, #2510) [Link] (1 responses)

I understand what you are talking about. I just never hit it in 33 years of being a Unix and Linux admin.

Users care about logging early kernel messages and containers? Those folks are sysadmins.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 22:02 UTC (Mon) by mjg59 (subscriber, #23239) [Link]

Your definition of user is overly limited. People who develop *on* Linux (rather than *for* Linux) are certainly users. And the increased focus on devops means that that includes a large number of people who are deploying large numbers of cloud instances, for instance. These are consumers of what we produce. They're people who want to care as little about Linux as possible. Making their jobs as easy as possible (such as by ensuring that all the relevant information they need to debug something is in the same place) is important in encouraging Linux adoption.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:16 UTC (Mon) by Zack (guest, #37335) [Link]

> We think that if we finally fix the system, then the users will come!

Nothing is getting actually "fixed" by systemd (unless it's 'fixing' users up to run not just Linux kernels but Linux kernels with cgroups enabled, and in all likelihood, pretty soon kdbus; Yay!)

It's about standardising and unifying the distribution landscape for easy deployment, which sound great in theory, except that it doesn't really solve any unsurmountable problems for Free Software deployment. Who it does solve an unsurmountable problem for is proprietary ISVs.

So yeah, the users might come, but not for the Free Software, and "Linux" will finally be on the desktop, but for all the wrong reasons.

Now if only systemd would actually fix some problems, instead of perceived and marketed ones ("bootspeed is important", "fragmentation is a problem", "Users hate UNIX", etc), it might still be a net gain, but so far it's mainly been a growing entangled pile of NIH, simply taking over existing functionalities.

So far it's been a research experiment to test the hypothesis "Those who don't understand UNIX are condemned to reinvent it, poorly" just to methodically find the exact answer to how deep the poverty can actually run reinventing it.

Anyone who criticizes systemd has a character flaw

Posted Nov 17, 2014 21:45 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link]

> Uh, the system took a few more seconds to boot. We had the dependency issue handled, not a problem. We had a bunch of more or less one-of-a-kind shell scripts where we didn't need them.
> Users never cared about this issue.
I'm a user of the initscripts and so far they:
1) Forced me to go into a datacenter at night because of an infinite loop in the BIND initscript.
2) Caused our clients to spend many many additional thousands of dollars because of various race conditions that usually happen in unreliable virtualized environment.

I say, that the classic Unix architecture is a crappy turd. It was designed for an era where computers typically had several running processes and very simple startup sequences. In the modern world it fails miserably and only systemd guys are actually interested in fixing it.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 14:20 UTC (Tue) by nye (guest, #51576) [Link] (2 responses)

>Users never cared about this issue. Developers did, either from a speed or architecture perspective.

Buh? Are you serious? As a user, I'm awaiting with gleeful anticipation the day that Debian releases a new stable with first-class systemd support because I expect it to file off a number of painfully sharp corners where I'm forced to interact with the init system against my will.

So far as I can see, the only people who are against it are *developers* who fear all change and want to keep everyone else from being able to enjoy it too.

This is why Debian has more pushback than other distros: it's an elitist distribution for developers. Whenever there is a decision to be made that involves a trade-off between ease of use for a small group of developers, and ease of use for normal people, normal people are likely to lose out - not because of any conscious decision to exclude them, but because the Debian community is so insular that it is practically in another universe to the majority of users.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 15:05 UTC (Tue) by Zack (guest, #37335) [Link] (1 responses)

> I'm forced to interact with the init system against my will.

Hi. I'm from the future and we received your message. I am sending this message to you, in the free past, to inform you that you can, as of *right now*, install systemd on Debian through the regular channels. Unfortunately, here in the future, we have no such options. D-bian is now mandatory and enforced by Skyne-d. It is now truly the Universal Operating system, but not quite as you hackers of the past envisioned it.

(P.S. We'd send back a robot to help, but we ran out of Legos.)

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 15:35 UTC (Tue) by nye (guest, #51576) [Link]

>you can, as of *right now*, install systemd on Debian through the regular channels

I have tried it out on a test system, but right now it's not really supported in stable and has enough rough edges that I'm not yet willing to make the switch on live systems. That's why I'm looking forward to first-class support in the next stable release.

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 0:15 UTC (Tue) by notninjaz (guest, #99725) [Link]

Bruce,

I would suggest giving this article a read:

http://artpetty.com/2010/07/08/6-steps-for-avoiding-group...

I think it captures the present situation pretty well and also offers ideas of what can be done.

Take care!

Anyone who criticizes systemd has a character flaw

Posted Nov 18, 2014 10:02 UTC (Tue) by OdyX (guest, #58768) [Link]

I find it incredibly irrespectful how you managed to hijack Colin's resignation article by pointing to an (unrelated) message from Russ for (yet another) generic rant about systemd, where nothing in Colin's message mentionned systemd.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 20:48 UTC (Mon) by damon.freer (guest, #99820) [Link] (19 responses)

Wow. Wikipedia tells me Bruce Perens was previous Debian Project Leader.

And this is how he behaves..

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 20:56 UTC (Mon) by BrucePerens (guest, #2510) [Link] (7 responses)

Sorry you don't like it. One has to stand up for what they believe in.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 21:17 UTC (Mon) by Tov (subscriber, #61080) [Link] (6 responses)

Indeed! But bringing your "soapbox to every funeral" is not the way to do it.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 21:52 UTC (Mon) by josh (subscriber, #17465) [Link]

> But bringing your "soapbox to every funeral" is not the way to do it.

...that is a disturbingly accurate analogy. Thank you for it.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 8:25 UTC (Tue) by Pc5Y9sbv (guest, #41328) [Link] (4 responses)

Rather than a soapbox at a funeral, please think of this as reflections at a wake.

I have been using and observing Linux since its inception, sometimes contributing bug reports or helping beta-test kernel and user-space systems, but mostly just using it full-tilt while I spent my own development time on other more specialized open source projects. My own professional development was deeply intertwined with Linux, but out of necessity also involved commercial contemporaries from Sun, SGI, IBM, HP, Convex, Cray, etc. I feel like I've been knee-deep in nearly every major Linux development from the first GPU support (2D and later 3D!), 64-bit (first Alpha and much later x86_64), SMP support, effective power management (idle, APM, ACPI, frequency scaling, tickless kernels, ...), HPC/clustering, virtualization (vmware, xen, kvm), automated provisioning, etc.

Most of that shared history felt like clearly motivated advancements, and well worth the risk and trouble for me to be an early adopter, give feedback, and promote the use of Linux in my own domain. By comparison, I've seen many recent (past 5-8 years) changes that really strike me as change for changes sake or of a train that has jumped the tracks. Changes that seem like big regressions both at a platform/architectural level and at a purely personal, selfish level of it breaking my daily workflow yet again. I've also noticed a pattern of dismissal as many regressions are characterized as necessary collateral damage rather than an indication that something is wrong with the development and QA culture.

I do not deny that developers should work on what interests them. I've spent my career doing so. But, I am saddened and a little bit frustrated by the zeal with which existing technology and abstractions are being destroyed. It is easy for me to understand why critics use loaded words like "cabal" to describe the situation. They feel that something of real value is being lost here and at a rate that defies normal bit-rot, and they search for an explanation. It really does feel like there is an agenda to co-opt and tear down what came before, as if it isn't enough to merely develop something else without stamping out the legacy options at the same time. The irony is that this is probably not due to specific developers like Lennart but rather it is the influence of the commercial interests finally leaking through. It feels like product managers making decisions for everyone, rather than the happy synthesis of a bunch of open source hackers collaborating in spite of their disparate management teams.

There is an uncomfortable changing of the guard that has been happening around Linux for quite some time. Perhaps it is just due to the natural lifecycle of such a long-term project, since participants are going through natural life phases as well. In the aggregate, first generation community members are not as active as we once were, and yet we can have a hard time relating to a new generation who never knew a world before Linux and who do not see existing solutions and compromises through the same lens of experience that brought us here. I hope we are just becoming crotchety old men and things are not actually falling apart. However, I fear that we are really environmentalists decrying the clear-cutting of our nature preserves in the name of progress...

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 20:09 UTC (Tue) by SiB (subscriber, #4048) [Link]

Thank you!

Colin Watson resigns from Debian Technical Committee

Posted Nov 19, 2014 4:33 UTC (Wed) by BrucePerens (guest, #2510) [Link]

Really nice writing. And I don't have a clue who you are. Do you mind if I use some of these points in another venue? Thanks!

Colin Watson resigns from Debian Technical Committee

Posted Nov 19, 2014 20:11 UTC (Wed) by tjc (guest, #137) [Link] (1 responses)

> In the aggregate, first generation community members are not as active as we once were, and yet we can have a hard time relating to a new generation who never knew a world before Linux and who do not see existing solutions and compromises through the same lens of experience that brought us here.

This is the reason I keep one foot in the Amiga community, despite the fact that the platform has not really been viable for decades.

Colin Watson resigns from Debian Technical Committee

Posted Nov 19, 2014 21:19 UTC (Wed) by halla (subscriber, #14185) [Link]

Well, the "lens of experience" is a really dangerous thing. It's way too easy to drop into the "we done that in the seventies, it ain't gonna work, no way" mode about things that might, superficially seem to be like something you've seen before, but, actually, aren't.

I've been coding since the early eighties and I've seen all the trends since then, and yes, I often feel that, oh gosh -- X is just Y, don't they ever learn. And then I discover that I was wrong. Even with applications and operating systems having live spans beyond twenty years or even more, the actual problems change, the actual user demands change, and software that survives has to change, too.

In short, experience is fine, but viewing the world through "the lens of experience" is exceedingly dumb.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 21:35 UTC (Mon) by mgb (guest, #3226) [Link] (5 responses)

> Wikipedia tells me Bruce Perens was previous Debian Project Leader. And this is how he behaves.

Bruce has done more for Debian than you or I are likely to achieve in our lifetimes.

And he is still working to make it better - in part by by protecting it from systemd's anti-innovation agenda.

Why are so many of those trying to force systemd into Debian not Debian users themselves?

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 22:27 UTC (Mon) by agrover (guest, #55381) [Link] (3 responses)

systemd has an anti-innovation agenda? I thought the opposition was due to too much, too fast innovation.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 22:49 UTC (Mon) by mgb (guest, #3226) [Link] (1 responses)

> systemd has an anti-innovation agenda? I thought the opposition was due to too much, too fast innovation.

I don't know of any software developer who is against innovation. Linux has become the world's most successful O/S ecosystem because the Unix philosophy enables and encourages innovation. And users pick the winning innovations.

Systemd is not new technology. Like most software it is a compilation of old technlogies. And that's OK. The problem with systemd is that it is a monolithic entanglement. Systemd's monolithic entanglement is the barrier to innovation.

When you look at why Linux won on servers, won on mobile, and failed on the Desktop, you see FreeDesktop.Org pushing bad designs like D-Bus and systemd.

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 23:25 UTC (Mon) by smurf (subscriber, #17840) [Link]

"Monolithic" my a**. systemd's public APIs are stable and well-documented, so what do you actually *want* that multiple repos would afford? What actually prevents you from cloning the systemd repo and just go off and innovate??

WRT dbus: what else would you want to use, given that dbus has pretty much won even before systemd showed up? Yet another incompatible RPC protocol? (And what exactly is bad about dbus anyway?)

Colin Watson resigns from Debian Technical Committee

Posted Nov 17, 2014 23:00 UTC (Mon) by dakas (guest, #88146) [Link]

Cobbling existing functionality together in one rather unstructured monolithic heap of everything is not innovation. And it does not become innovation when done in a rush either.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 20:22 UTC (Tue) by HelloWorld (guest, #56129) [Link]

Yeah right, systemd is definitely anti-innovation.
http://www.youtube.com/watch?v=ExWfh6sGyso

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 3:08 UTC (Tue) by ghane (guest, #1805) [Link] (4 responses)

And if I turn it around,

> This is how he behaves.
> Wikipedia tells me Bruce Perens was previous Debian Project Leader.

shouldn't I be listening to him? He may be wrong, or senile, or rabid, but some respect in _listening_, before we ignore his ramblings, is not unreasonable.

I am not saying his experience (say hi to Steve!) has made him smart, but I can profit from his experience (or anecdotes).

(I installed the systemd PPAs on Ubuntu 14.10 pre-release 6 months ago, and it did break boot a couple of times during updates, but that is neither here nor there. I still haven't figured out journolctl, systemctl, etc, but as a user, I can live with either option)

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 7:06 UTC (Tue) by torquay (guest, #92428) [Link] (3 responses)

    shouldn't I be listening to him? He may be wrong, or senile, or rabid, but some respect in _listening_, before we ignore his ramblings, is not unreasonable.

This is an argument based on appeal to authority, which is ultimately a logical fallacy in this context.

Based on Bruce Perens' recent comments that are clearly uneducated, short-sighted and factually wrong, his authority and/or credibility is rather questionable. Instead, he comes across as someone stuck in the past, unwilling to see today's problems, and today's solutions to these problems.

Is it fallacious to suggest listening respectfully?

Posted Nov 18, 2014 10:15 UTC (Tue) by pjm (guest, #2080) [Link] (1 responses)

To say "Bruce has relevant experience, therefore Bruce' position must be the correct one" would be a logical fallacy. That is not what ghane said: indeed, ghane was most explicit that Bruce might be wrong, and that we might take a different conclusion from Bruce.

ghane merely suggested paying attention (not being too quick to dismiss), and being respectful. When making decisions that involve guesses about the future, there's nothing unreasonable about listening to people with relevant experience — nor about respecting others. (I'd rather not use this thread to discuss whether Bruce is a person with relevant experience, authority or credibility.)

Is it fallacious to suggest listening respectfully?

Posted Nov 18, 2014 11:58 UTC (Tue) by ovitters (guest, #27950) [Link]

IMO it is easier to first tell the person to behave respectfully and explain themselves again. That makes it clear that the current way is not ok. Once that is established, then go on to other things.

FWIW, I think some of the communication from systemd project could be improved. However, this post/article is about Colin leaving. Not much to do with systemd, so rather poor to make it about that.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 10:40 UTC (Tue) by sdalley (subscriber, #18550) [Link]

Very clever. But the trouble often with being very clever, is that it tends to make one arrogant, and disinclined to respect others whom one thinks aren't as clever as oneself. The people you don't respect, you don't listen to properly, and you becomes deaf to where they do have a valid point or experience of life.

Just sayin'.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 6:38 UTC (Tue) by stefanor (subscriber, #32895) [Link] (2 responses)

Also relevant: https://lists.ubuntu.com/archives/ubuntu-release/2014-Nov...

And of course the linked blog post. It's not all about systemd, folks.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 17:47 UTC (Tue) by drag (guest, #31333) [Link] (1 responses)

There seems to be quite a few Debian developers leaving Debian for reasons only superficially connected to Systemd.

Found another one; http://lists.alioth.debian.org/pipermail/pkg-systemd-main...

This should be a huge warning to the people involved in the management of Debian. I like Debian a lot and it's been around for a very long time, but Debian dying wouldn't be the first time 'the commmunity' did a end run around a badly performing bureaucracy for the sake of improving a project.

The most notable example of this I can think of is the move from 'XFree86 Project' to 'Xorg'.

If this sort of thing continues (say for the next couple years) it may be time for all the other distributions and businesses that depend on Debian to start looking into setting up their own collaboration organization. I have the hope, however, this is just a series of 'end of year lifestyle re-adjustments' being announced by different individuals for personal reasons, it is not systemic, and this whole thing will just blow over.

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 19:51 UTC (Tue) by mathstuf (subscriber, #69389) [Link]

See: https://lwn.net/Articles/620878/

Colin Watson resigns from Debian Technical Committee

Posted Nov 18, 2014 23:15 UTC (Tue) by julian67 (guest, #99845) [Link] (6 responses)

I'm a long time non-subscriber, non-member but regular reader of LWN. I finally created an account to add some comments. Here goes:

Thank you Bruce Perens, for remaining civil and for making a reasonable case and continuing to do so while it seems your rather easily parsed sentences are wilfully misinterpreted, misrepresented and misconstrued, all while you are subject to being sluiced with enough mud to create a planet, apparently because some of it is bound to stick.

I am a Debian user. I'm not a sysadmin or CTO of some fabulous tech enterprise, or an employee of anyone with a dog in this fight, just a regular user who has a Debian Stable server and various Debian Testing desktops and laptops. This means I'm running an olde worlde sysv-init stable server and also machines which use systemd. I don't have an allergy.

What I really dislike falls into two categories:

first is the brazen hypocrisy of posts such as https://lwn.net/Articles/621426/ where Mr Perens is vilified for supposedly committing the crime of the logical fallacy of "appeal to authority" *accompanied* by a slew of personal insults ("..uneducated, short-sighted ... credibility is rather questionable.... stuck in the past, unwilling to see today's problems..)

This is the grossest hypocrisy. But I suppose the invective and poison might serve to distract from the fact that BP was not advocating an "appeal to authority", but merely suggesting that life experience can bring insight and might be regarded as a valuable resource, not just a ghastly and unavoidable mechanism for the manifestation of grey hair and piles and ulcers (can you tell I'm not young?).

I don't think it's disgraceful that BP is articulating his concerns in a civil and reasoned way in places and in situations that are absolutely relevant. I do think it's disgraceful that a handful of people who apparently are not even Debian users, let alone developers, bark like dogs at him whenever he does so. Four legs good, two systemd legs better.

The other thing I really dislike is software that is designed to break other software. If people want/need to create an init system that will, by design or knowing indifference, damage or coerce other free software projects (i.e. the BSDs) then how can that software be the default in Debian? If it fits Fedora or RHEL or Suse, fine. But Debian is not Fedora, it is not RH and it is not Suse.

Colin Watson resigns from Debian Technical Committee

Posted Nov 19, 2014 13:35 UTC (Wed) by mstone_ (subscriber, #66309) [Link] (5 responses)

"As you can see, I found out this morning that anyone who criticizes systemd has a character flaw. It's not a technical or policy argument at all. You can't really even start talking about it without being criticized for the tone of your discussion.

This is mad, folks."

This was civil and made a reasonable case? You must define that differently than I do. That post was a classic troll. It added nothing substantive to the discussion, obliquely criticized some set of people, and invited them to make a similarly content-free rebuttal.

Colin Watson resigns from Debian Technical Committee

Posted Nov 19, 2014 17:19 UTC (Wed) by namreh (guest, #99869) [Link]

Thank you Bruce for your hard work over the years and trying to keep Debian on track. It sure seems like an uphill battle at the moment.

Please keep to the high road and ignore the invective that keeps bubbling up out of the bitrot in the bilge tubes of the internet.

Yes, it is civil

Posted Nov 19, 2014 19:40 UTC (Wed) by BrucePerens (guest, #2510) [Link] (3 responses)

That was my exact observation of the state of the discussion.

Now, I am sure that systemd folks have had no lack of vitriol in their direction. But it was and remains my observation that there is a juggernaut of software change presented by the systemd team, some of which has merit and the rest less. And it's perfectly valid to stand up and say "do we really need this?". But what I got was criticism = hate, criticism = bad behavior, criticism = oh what a pity that Bruce has gone so low.

OK, I have had any number of people attempt to use emotional tactics to demotivate me, and I pretty much know what to do by now. Which is mostly ignore it. And I got what I needed for this discussion, which was all of the speaking points for an editorial in another venue. The people have been polled and some of them were even worth representing.

And yes, I still think it's mad.

Yes, it is civil

Posted Nov 19, 2014 19:55 UTC (Wed) by mstone_ (subscriber, #66309) [Link] (2 responses)

Wow. If you have something substantive to say, just say it and stop acting the martyr. This might be a really great time to just start a new distribution, as the debian developers have made it pretty clear that they want this time-wasting nonsense to stop. We'll have two init systems for jessie. If people keep maintaining both of them, we'll keep two init systems after jessie. If one turns out to not be useful it'll get dropped. We can just deal with it, without the drama. It seems as though you've said what you wanted to say, and have convinced those that you are likely to convince. What's the point in continuing to pick at it? If you're right, the project will dump systemd at some point in the future and do something else. If you want to help make things better, find a way to contribute. If you just want to complain, start a new distro built around whatever it is that you stand for these days that you think the debian developers are too stupid to figure out.

For the record, I hated systemd also, but now I don't want to be associated with the anti-systemd crowd and would rather support it than just sit and whine. Some of my machines have systemd, some have sysvinit.

Yes, it is civil

Posted Nov 20, 2014 3:34 UTC (Thu) by namreh (guest, #99869) [Link] (1 responses)

I think that one of the problems with computer science in general and FOSS especially, is that a substantial number of contributers and users are mentally ill / unstable. This makes it hard on the rest of the contributers and users who have to accommodate them.

Refactoring and improving the init system is the easy part.

Yes, it is civil

Posted Nov 21, 2014 17:27 UTC (Fri) by tfx2 (guest, #92047) [Link]

Please don't blame mental illness for bad behavior. There are plenty of people with mental illness who are not engaging in the harassment of people associated with the systemd project.

Accommodating people with mental illness doesn't mean that you have to accommodate toxic behavior, and eliminating toxic behavior is actually a pretty good accommodation. If harassment and trolling was far less common, we would have greater participation in open source, and it would remove one source of stress that contributes to burnout.


Copyright © 2014, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds