Irrationality at it's best

Posted Oct 20, 2014 0:30 UTC (Mon) by josh (subscriber, #17465)
In reply to: Irrationality at it's best by jch
Parent article: The Debian init system general resolution returns

I'm suggesting that if half the energy that went into telling other people to maintain sysvinit compatibility or otherwise complaining about systemd (across all the various people doing so, not just the GR proposer here) instead went into writing and submitting patches, they'd have nothing left to worry about.

A GR won't make support magically materialize, and if such support already exists, a GR isn't necessary to obtain contributions from it. So, this GR is either futile or redundant, and either way it's a giant waste of everyone's time.

Irrationality at it's best

Posted Oct 20, 2014 2:13 UTC (Mon) by mgb (guest, #3226) [Link] (18 responses)

It is not about patches. It's about protecting Debian from deliberate breakage.

Everything in Wheezy works without systemd.

Anything in Jessie which now depends on systemd was deliberately broken.

Irrationality at it's best

Posted Oct 20, 2014 2:35 UTC (Mon) by josh (subscriber, #17465) [Link] (17 responses)

To repeat what I've already said elsewhere in this thread:

Such software is not "broken"; it works just fine with its dependencies (systemd, etc) installed. If you want it to work with other init systems, that is a wishlist feature request. The maintainer, or upstream, may choose to implement that feature request for you. You or someone else could also choose to submit a patch, and for the record, I do think maintainers should accept reasonable and maintainable patches.

The interesting case comes up when no patch exists or is forthcoming. The proponents of this GR would have the packages removed or otherwise held hostage to force the maintainer to add support for sysvinit, with no provisions for software that actively uses and depends on systemd features.

To give a parallel example: suppose you want to build a Debian system without glibc. You could attempt to make piles of software work with another libc implementation, and you'll sometimes run into bugs where software depends specifically on features of glibc. Such software is not broken for doing so; it simply avoided reimplementing glibc functionality to work with other libc implementations that don't have that functionality. You could submit patches for such projects, to make them build without glibc; some projects might take those patches. However, the bugs you might file are wishlist at best, and a maintainer would not be particularly unreasonable to close such a bug as wontfix if the patches needed to work with another libc are large, invasive, and not easily maintainable.

Irrationality at it's best

Posted Oct 20, 2014 3:30 UTC (Mon) by mgb (guest, #3226) [Link] (15 responses)

It worked before without systemd. It doesn't work now without systemd. It was broken. Deliberately.

Debian doesn't need patches. We have the Wheezy source we can recompile.

We just need a GR to prevent systemd proponents from deliberately breaking things again.

Irrationality at it's best

Posted Oct 20, 2014 3:41 UTC (Mon) by josh (subscriber, #17465) [Link]

In short: you claim it's always a bug to depend on systemd, and you completely ignore the possibility that systemd has useful functionality that other software may not wish to duplicate. (You're also ignoring the possibility of new software appearing in Jessie which never had support for other init systems in the first place.)

You seem to be either clue-resistant or deliberately ignoring my responses, and I have no further interest in arguing this further with you. I believe anyone reading the comment thread can see the point we're each trying to make, and evaluate each for themselves.

Irrationality at it's best

Posted Oct 20, 2014 7:39 UTC (Mon) by tao (subscriber, #17563) [Link] (2 responses)

My old 386 used to work with older versions of Debian. It doesn't work now with modern versions of the kernel. It was broken. Deliberately.

Debian doesn't need patches. We have the Woody source we can compile.

We need a GR to prevent the proponents of recent versions of the kernel from deliberately breaking things again.

Irrationality at it's best

Posted Oct 20, 2014 9:51 UTC (Mon) by josh (subscriber, #17465) [Link] (1 responses)

Nice. Here's another:

It worked before without glibc. It doesn't work now without glibc. It was broken. Deliberately.

Debian doesn't need patches. We have the Bo source we can recompile.

We just need a GR to prevent glibc proponents from deliberately breaking things again.

Irrationality at it's best

Posted Oct 22, 2014 7:34 UTC (Wed) by jordi (guest, #14325) [Link]

Brilliant! :D

Irrationality at it's best

Posted Oct 20, 2014 13:23 UTC (Mon) by jrn (subscriber, #64214) [Link] (10 responses)

Is this a hypothetical or do you have an actual package and bug in mind?

Irrationality at it's best

Posted Oct 20, 2014 13:57 UTC (Mon) by josh (subscriber, #17465) [Link] (9 responses)

I'm not sure whether you intended that question for me or for someone elsewhere in the thread. Could you clarify which possible hypothetical you mean?

Irrationality at it's best

Posted Oct 20, 2014 14:10 UTC (Mon) by jrn (subscriber, #64214) [Link] (8 responses)

The comment I was replying to, by mgb: "It worked before without systemd. It doesn't work now without systemd. It was broken. Deliberately."

I was hoping someone could clarify what was broken, so it can get fixed in time for the release.

Irrationality at it's best

Posted Oct 20, 2014 14:17 UTC (Mon) by jrn (subscriber, #64214) [Link]

... if it is actually broken.

Especially when people disagree a lot and their identity seems to be wrapped up in a question, actual practical problems can be easier to solve than finding abstract principles to agree on.

Irrationality at it's best

Posted Oct 20, 2014 14:39 UTC (Mon) by josh (subscriber, #17465) [Link] (6 responses)

Ah, sorry; lwn's comment threading is not ideal.

Irrationality at it's best

Posted Oct 20, 2014 14:52 UTC (Mon) by johannbg (guest, #65743) [Link] (5 responses)

As far as I can tell commenting is not limited to subscribers either which arguably it should be.

Irrationality at it's best

Posted Oct 20, 2014 15:03 UTC (Mon) by Cyberax (✭ supporter ✭, #52523) [Link] (4 responses)

Ah, but everyone knows that only supporters can write good comments!

There is a ton of insightful guest comments here, so banning them is counter-productive. However, banning all the systemd-related threads might make sense. The SNR has become way too low in these threads.

Irrationality at it's best

Posted Oct 20, 2014 15:10 UTC (Mon) by johannbg (guest, #65743) [Link]

Less about keeping the comments down more about having those comments of questionable quality, support the site so we can get good article in exchange of having to read all those.

Irrationality at it's best

Posted Oct 22, 2014 19:48 UTC (Wed) by Kamilion (subscriber, #42576) [Link] (2 responses)

Personally, I've been running statistics in my head whenever one of these 'RAWR' topics come up; and to be honest, LWN is where I come to view generally courteous digressions on a particular poster's opinions and experiences. Often I get wonderfully pertinent data on a wide range of topics (see the 'filesystem geek' above outlining his experiences with btrfs -- mine differ!) that I care about.

I don't have much money, but I do keep my LWN subscription running to pick up some of the most meaningful insight into the linux world and it's players.

So far, I think Raven667 summed it up best with:
"If things were as bad as the detractors say it would have been obvious for a long time and they wouldn't be talking about 'potential' problems, there would be a long line of angry pitchfork wielding sysadmins."

So far most of what I've heard has been grunts of indifference, a couple 'oh thank god I can finally kill that horrible initscript hack', and only one outright dismissal of systemd from a hardcore gentoo'er I know.

But hey, I helped out Gerard Beekmans and Jessie Tie-Ten-Quee with the getting the first copies of Linux from Scratch out the door -- I understand my system enough to know why things are. I understand others may not necessarily have that clarity, nor do I fault them for not wishing to develop it. Humans are by nature self-specializing, after all.

Irrationality at it's best

Posted Oct 23, 2014 21:13 UTC (Thu) by cdmiller (guest, #2813) [Link] (1 responses)

Well, there was an interesting systemd discussion on the LOPSA freenode channel last week with some views from sysadmins. A systemd developer there was initially pretty defensive towards any criticism.

Around here a few consternations expressed by sysadmins are the implementation of the binary log (corruption handling, reinvention of the wheel), lack of log shipping (yay let's run 2 loggers now), and a perception of "crazy feature creep" (one current joke/betting pool is when windows-registryd will appear).

In the server realm a couple precepts are stability generally trumps features and "sudden or quick" forklift changes to services are to be avoided if possible. If or when systemd becomes a problem or too much of a pain point our sysadmins will eventually route around it is the sysadmin feedback I receive at present.

Irrationality at it's best

Posted Oct 23, 2014 22:19 UTC (Thu) by johannbg (guest, #65743) [Link]

Sysadmins on these parts like it anyway everybody is just waiting for General Resolution: init system coupling [1] voting so DD needs to just get this over with and move on.

1. https://www.debian.org/vote/2014/vote_003

Irrationality at it's best

Posted Oct 20, 2014 21:57 UTC (Mon) by rodgerd (guest, #58896) [Link]

Given the wording it would be entertaining to start filing bugs for packages not working right with other random init systems, such as the HURD's native system, or openrc or what have you. All of which would be considered blockers under the wording of the GR.