Leaving Debian
corecursive.com
corecursive.com
The project does, however, sometimes have its share of people who dogmatic/pedantic to the point where it really doesn't help the cause. This might be expected from something that's a passion project for many of its contributors, and to Debian's credit, they've always been able to resolve these conflicts eventually.
The article mentions the systemd drama. The recent kerfuffle around the 'which' command might be a smaller example of this kind of thing: https://news.ycombinator.com/item?id=29026623 And note: As silly as that discussion was, when it got widespread attention, it was actually already resolved.
To be honest, Desktops reached their peak for me with Gnome 2. I'm using MATE and I'm good.
Same; I've yet to find anything more stable and well put together.
I didn't follow this at all at the time and thought it was something like 90% for. Turns out, it was closer to 50/50:
> The vote came down much as expected, with a 4:4 split between systemd and Upstart proponents. Anthony Towns analyzed the votes and declared a tie between systemd and Upstart, which left it up to the chairman to decide by using the casting vote. Garbee did just that, voting for systemd, which makes it the Debian Linux default init for jessie.
Having a more active & ongoing paradigmatic discourse would have been interesting.
So I kind of want a synopsis for upstart. These threads might provide some of that material. I'd touched upstart a couple times back so long ago, & the words "event-based" keep flashing before me, but I don't have much recall on upstart. What was it bad at? What didn't people like? What did people like? What was novel & interesting? What had potential but didn't quite work out? What was the best operational parts of it? There's so much tech in the world that doesn't quite make it, or which did make it & fell (CORBA), & we really need more Speakers For the Dead, people who can try to tell us the story of the tech earnestly, who can help us memorialize something by drawing together as much understanding, from as many points of view as they can.
The author of systemd addresses this question in the original blog post which announced the systemd project (look under the heading “On Upstart”):
Upstart is still used by ChromiumOS: https://www.chromium.org/chromium-os/chromiumos-design-docs/...
A lot of what you find in Debian also comes from what Ubuntu does. In fact, I believe that many people hired by Ubuntu in the early days were experienced maintainers for Debian.
20 years ago, the problem was all about getting Linux consumer desktop friendly. On the server side, it was mostly OK, people there were not afraid of a TTY, because it was the primary entry point to learning Linux, and Unix in general, whereas in the Windows world, people first approach was (and still is) the GUI. Debian wasn't as good as it is today in the desktop area.
Ubuntu appeared on the market mainly to solve that problem (they were not the first ones to do that), and "leverage" on the acquired expertise to make money.
Nowadays, the money is in servers and IoT, so they are focusing to that segment of the market.
The nature of open source meant that of course, Debian benefits from their work.
The systemd drama of a few years back really soured me on debian though. On the plus side, it has made me explore OpenBSD and FreeBSD, both of which I use these days. Also I'm using Devuan where I need or want Linux.
Still, I miss having the old debian which eschewed trendiness and valued stability more.
Solution: switch to antiX, MX Linux or Devuan.
This may be wrong but it feels like people were stuck in a local maximum and don't realize that if they move more than a little from where they were, then they'd actually be in a better place. On the other hand, solving again a problem you 've already solved, is not the most enticing part of the linux experience ;)
In other words: I'm not a fan on ideological grounds but from a practical standpoint it seems to work just fine.
and to address this new concern, i dont think replacing systemd would be so hard. and i dont think it would be so difficult to replace systemd init/pid1 while still using other systemd services. systemd is more a mono-repo; there's very little notable hard runtime coupling between the different pieces.
i expressed elsewhere in the comments a sadness that upstarts death has lead us to monoculturism. i do want to see us exploring & trying. but i dont think systemd init/pid1 is deeply rooted, i dont think it's a particularly difficult thing to replace. some not-very-big translation layers will get through 99.9% of what distos actually use, to address point 1. and i simply disagree on point 2..there already are competing offerings and what systemd kffers does not often insist on being run under systemd.
I did. I'm still not happy with the way it was forced in, and how the predictions of teething pains of a from scratch init system being developed by inexperienced developers came true. I still remember the growing pains from PulseAudio, the previous project by the same developers.
But SystemD was a response to the definite problems of past init systems. It might not be the optimal UNIX philosophy architecture, but it's grown into something stable and powerful.
At the end of the day, as a working IT or SWEng professional, you just want to get shit done, so you learn a new abstraction for the zillionth time, put it to practice, and get on with the problem you were trying to solve. Having moved from MS-DOS, to OS/2, to Slackware, to RedHat, then Debian (and staying on Debian for decades, for desktop and server), I trust the Debian devs to put out the most stable and reliable OS I've ever seen.
It's just that we all have different ideas about what that is.
daemontools-style supervisors are what I prefer, myself. I also acknowledge that there isn't a clear winner in that space.
What I object to are people enforcing systemd by making it a requirement for other projects.
This in a nutshell is why all progress is always a god awful painful mess. No matter how modest.
I used to be self-proclaimed advocate for change. Technological, organizational, political, all of it. Everyone marching arm-in-arm towards our glorious future.
Now I'm happy when any one manages to change anything at all. I no longer even care how good or bad an idea is. Better to try and then learn from failure than never try at all.
Eh, it's been a while now, though, and I guess most of that has gotten worked out.
I knew of joeyh from hn and a previous hn post is how we ended up talking[1].
The episode is a lot about the experience of being in the Debian community in the early days and the experience of being a free software developer, or at least that is the part of of his story that was most interesting to me and that I tried to highlight. How communities form and how they build up a shared understanding and culture is something I'm trying to understand.
I just decided to avoid the friction and not bother publishing my .debs.
I got my key signed with the same simple procedure. The only troublesome part was to find Debian devs to meet physically, because there were only 3 in where I live (Hong Kong), and only one of them had a valid key (others have expired or just lost keys). Two signatures are needed, so I needed one more. Luckily, I happened had to travel to the US for work, so I solved my problem.
1. go ahead and pull that Ethernet cable, your precious networked daemon will restart thus having to lose whatever cached or learned data. This is a major problem for designer of custom daemons.
2. Try and interface with Juniper DHCP servers at major ISP using systemd-networking and /etc/systemd/network settings. You’ll be forced to revert back to ISC DHCP server and/or NetworkManager.
3. Bootup and actually making your IP-address-senstitive daemon croak because your custom Ethernet NIC requires ‘ip addr/link up’ in /etc/rc.local
4. Bootup and get to network-online state only to find out that your netdev really hasn't gotten a dynamic IP address yet, your networked daemon that requires those IP address to be defined on the netdev will croak if not handled properly.
5. You cannot firewall against PID 1.
For UNIX box that are true server or not cloistered behind an ISP-provided gateway/router, that was more than enough to ditch systemd and go Devuan distro.
You cannot operate without systemd in Debian 11+. It’s now a REQUIRED “feature” and a veritable albatross around Debian’s neck for those serious server needs.
Installing openrc doesn't do that, but also doesn't remove any systemd components, so unclear what would happen if I tried to do that.
you just cannot unplug systemd 100% thereby contaminating the disk with useless stuff.
If you could mention these services that require it, we could look at the services and see what needs to be done.
cannot uninstall them.
Most of the issues you raise sound like you simply didn't want to investigate in depth even though you knew you had an exotic setup.
> I was perfectly happy with systemd myself. I did empathize with people who didn’t like it for whatever reason, but it was clear to me that every Linux distribution except for one or two would be switching to systemd eventually. And I felt that I was pretty good at prognosticating, this kind of thing. So I was, if Debian can’t make the obvious decision without all this much drama, it’s too much for me.
His way is thinking is basically the same as of all Systemd supporters. "I see no problem, it is obvious, everybody is doing it so we will do it sooner or later, anyway."
But this is not, and never has been the point. The crux of the matter was very technical[0]: by gluing GDM to Systemd, the Systemd developers made it very difficult for distributors to support other init systems. They basically forced everybody to switch, and Systemd was not bug-free - it gave sysadmins around the world quite some work, and in some cases, also caused unforeseen problems.
In hindsight, we can see it was unnecessary. Devuan should never have to exist. Debian is offering some support for System V init now. All this drama was completely redundant if someone had not made one controversial architectural decision. Yes, we would have probably transitioned anyway - but on our own terms and at the time we could choose. And Joey would probably be part of Debian still.
[0] Maybe it was also political, but this is beside the point and doesn't change anything.
This is a pretty clear indicator right here that influential people weren't listening on the issue. Any time you state "for whatever reason" is a clear sign you don't care for their reasons.
Is the existence of Devuan not a good example of a conglomeration of reasons so that I don’t have to spend time doing research for this request you probably don’t want answered? You’re still pushing the reasoning aside and saying “those reasons weren’t good enough so we put them in the whatever catgory”. I understand it’s often easier to hand-wave away well after the fact and the damage has been done. The Debian community didn’t move on from it, it split.
All hundreds of them. That was debated ad nauseam both here and on practically all Linux websites. See https://ihatesystemd.com/ or https://nosystemd.org/ for some examples.
So does every other version of Linux. If this enough to dissuade that there is no difference of opinion or difference in values and process handling and isolation then I cannot help you on this topic.
Edit: Also, the proliferation of Linux distributions is illustrative of the same problem I think. There are a lot of distros out there that seemingly cannot really justify themselves, I've been using Linux for 20 years and for some of these distros it's still really hard for me to figure out what the difference is or why anybody would use it without spending days digging around and analyzing patchsets in random git repositories somewhere. Now just imagine how utterly incomprehensible this is to a non-technical person who doesn't have a clue what an init system is...
From the homepage:
> Devuan GNU+Linux is a fork of Debian without systemd that allows users to reclaim control over their system by avoiding unnecessary entanglements and ensuring Init Freedom.
From the Init Freedom page:
> Init Freedom is about restoring a sane approach to PID1 that respects portability, diversity and freedom of choice.
> Other Free Unices Without systemd
> One of the critical arguments against systemd is the lack of portability to other Unices and the possibility that Linux-based systems using systemd would detach from the UNIX world. OpenBSD, FreeBSD and other BSD OSes are de facto incompatible with systemd or rather, systemd is incompatible with Unices not running Linux.
Also, the statement about portability is weak, as someone did manage to port systemd to BSD: https://github.com/InitWare/InitWare
For me, in general it just seems really difficult to justify creating an entirely new Linux distribution because you're waiting for somebody to port a few packages to BSD. That logic doesn't really follow. That's what I meant before, even the Devuan home page has some really hand-wavey and confusing statements.
Whether the results are the same is irrelevant to the difference in approach.
Edit: I am also even more confused as to what that has to do with init systems because BSD never used a system V style init, whereas Linux distros typically did for that period before systemd. So this stuff wasn't portable anyway. See what I mean that the logic doesn't follow? I really can't understand what someone meant when they wrote that.
To answer this question. This is about base compatibility so ports don’t apply. The ports collection has to be installed separately, hence incompatibility.
So in that sense, systemd is helping retain some diversity in Unix like systems...
I'm sorry but I don't buy this, I think you and those sysadmins missed something. The situation with consolekit was so bad that something had to be done about it, if you were blindsided by it, you probably weren't paying attention to that area until it was too late. That's unfortunate but it doesn't really make sense to blame the systemd developers for "making it difficult" when they were the only ones trying to actually solve the problem at that time. IIRC the upstart developers never made progress on this.
Surely he can look back over the last few decades and feel proud of his work and the benefit it has brought to so many users. The tech treadmill can cause us to lose focus on the things that really matter or that really inspired us to begin with, so this is a refreshing perspective.
"datePublished": "2021-11-02 00:00:00 +0000"
Unfortunately, this kind of mistake on dates is usually a plain lie to fool the SEO and get more visits. It's so frequent it's not even surprising.
^F whichI am very happy with OpenSuse.
The discussion went approximately:
"`which` is nonstandard and doesn't always work the same. Maybe we should remove it and have everybody use `command -v` instead, which is standard. Also, in testing, I've added a deprecation notice to prepare for it."
"Let's have a vote to see. A lot of people use `which`, and that deprecation notice breaks some builds.
...
Vote says no. Remove the deprecation notice and leave `which` where it is for the next release"
Later, on Hacker News: "Debian might be removing "which"! Why are they doing this?"
Most of the commenters, it seems, never actually read the entire story.
In conclusion: Debian couldn't care less about their users' time. No respect for the users.
I'm trying to make my real thoughts on this palpable. Idiotic doesn't even begin to cover it.
That which(1) is not portable is utter nonsense.