systemd/Devuan

Home Forums Linux systemd/Devuan

This topic contains 6 replies, has 3 voices, and was last updated by  RChandra 3 weeks ago.

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #5766

    xcal99
    Participant

    Just wondering what anyone’s thoughts are on the systemd dustup that’s gone on during the past couple of years, and if anyone has tried the resultant Debian fork, Devuan?

    #5767
    Christian
    Christian
    Participant

    systemd tries to replace everything

    I run Debian on my desktop and all my servers but I’m not involved in its development. So I had to trust the community to make the right decision between staying with sysvinit, or choosing between upstart and systemd.

    I had most of the same concerns that I think led to the fork; that systemd’s developers were willing to screw over BSD by promoting tight coupling between Linux and popular desktop software like KDE and Gnome. That systemd was bloated and tried to do too much (the above gif).

    Fast forward a few years and it seems Gnome still works on BSD and I’ve mostly gotten used to systemd. I’ve written a few Unit files. I haven’t explored many of the *extra* features and still prefer doing things the old way with cron, autofs, syslog, etc.

    There does seem to be some problems when using other software that depends upon cgroups such as Linux Containers. I was surprised to learn last month the Jessie didn’t support unprivileged containers (but they’ve just released Stretch so we’ll see).

    Also, it was my impression that one could still use sysvinit in Debian if desired. I haven’t tried Devuan, but only because I haven’t tried any new distros in a while. The only other one I’m currently using is Arch on my Macbook. Arch went even further with systemd (no cron and no syslog by default). Debian at least has tried to maintain backwards compatibility for the init scripts and for other tools / services.

    #5768
    Christian
    Christian
    Participant

    I”m wondering if we’ll be going though the same thing in a few years with X11 and Wayland.

    #5771

    xcal99
    Participant

    Yeah, things get interesting when they start calling something you’re used to “legacy” software.

    Over the past 4 or 5 years I’ve used about 20 Linux distros and have wound up liking the Debian based ones best. On my desktop machine I’ve settled in with Debian and LMDE – on my laptop I use Bunsen Labs (‘Arch Bang Reborn’ as they say) but Debian is the workhorse of the three. And yup, there have been a few glitches, but nothing horrendous.

    The project’s description of Stretch (Debian 9) includes salient comments like:

    “…the default MySQL variant is now MariaDB.”
    “…the X display system no longer requires root privileges to run.”
    “…The Stretch release is the first version of Debian to feature the modern branch of GnuPG in the gnupg package.”

    Systemd 232 is just one item in a long list of included updated packages.

    They’ve dedicated Stretch to the memory of Ian Murdock. When I was reading the arguments over systemd I was kind of leaning towards the side of those objecting to it, thinking Murdock would have taken that stance as well, but then I remembered his essay Do operating systems still matter? Even Torvald’s response seems to be a shrug of the shoulders. I guess it’s just us “legacy” types who get nostalgic about something like sysvinit.

    Any way, I have downloaded Devuan 1.0.0 and burned it to a disc. Have only run it live briefly, but I’ll be loading it onto my lap top. If nothing else it’ll be interesting to watch how it holds up as systemd moves forward.

    #5772

    xcal99
    Participant

    Correction: BunsenLabs calls their distro “Crunch Bang Reborn”, not Arch Bang.

    #5799

    RChandra
    Participant

    What I hate is Poettering’s in-your-face, you-adapt-because-we-don’t-have-to attitude. A good example is when he decided to usurp a long-used kernel command line argument for systemd’s own purposes. This caused some people to have their consoles spammed with systemd messages, and no particularly good way to get that not to happen. IIRC, the kernel developers’ response was to include kernel code which checked which process was trying to do certain things, and deny it to anything called systemd due to this decidedly non-community behavior.

    Another example is their stance on programming. As an example, their attitude is your network daemons ought to be more robust and resilient to not having network, instead of doing things like making network services depend on the “network facility” finishing its startup. While Poettering sort-of has a point, it’s mostly an edge case, and bloats every network service by having it have all sorts of “does the machine have an address now?” code and similar, whereas 99.3% of the time, just waiting for the network startup to complete would be quite sufficient.

    My current problem is NTP rarely starts because of this. I have to add scripting to /etc/rc.local to check if it’s running, and if not, start it.

    #5800

    RChandra
    Participant

    My current stance is, there is little evidence to suggest anything other than eventually the whole Linux community will adopt systemd, so I might as well just learn how to use it. For the vast majority of day-to-day things, I’m proficient. It’s on balance not all that bad operationally. In philosophy, it’s quite bloated, and is very anti-Unix (do one and only one thing well, and provide easy-to-use interfaces, usually text on input and output (piping)). It makes Microsoft’s embrace and extend with standards look like child’s play. Still, it does provide a lot of useful facilities.

    What really gets nuts is when something doesn’t work, despite all the friendliness of things like journalctl help messages, it can be maddening to try to figure out why something isn’t working or happening. As an example, I was sitting there with an unbootable system, in the emergency recovery shell, trying to mount something. Well, systemd is king of the hill, nothing, including mounting, happens without its express permission. The mount(8) command wouldn’t fail, it’s just that as soon as that was completed, apparently systemd was saying, “no, you don’t; umount”. All I needed was that one thing (I forget, could have been partition, maybe volume group lv) mounted to fix a config file on it, but the system wasn’t having anything of that. I think I eventually booted a live CD to edit that one file, and things were fine.

    Just…some things are really stupid that way. Upstart was quite arguably much less of a pain. And a personal pet peeve, all the descriptions are capitalized, when there’s no reason (like “Getty On Console %i” or somesuch, instead of just “getty on console %i”).

Viewing 7 posts - 1 through 7 (of 7 total)

You must be logged in to reply to this topic.

Comments are closed.