Debian Init System Discussion Is Still Unsettled

[andrebrait]

You're forgetting Debian jessie, for a stable bleeding edge distribution. Never mind that you can also use pinning with Debian apt, to keep your base system as the stable distribution, while cherry-picking packages from testing, unstable, and experimental for installation. Works beautifully, and is a great way to keep your system rock stable, while enjoying the latest versions of whatever you need. That's something Fedora cannot do, from my knowledge.

idk about debian, but when I use "testing" ppas in Ubuntu I end getting weird things. I installed GCC 4.8 from the compiler toolchain ppa and I ended up with a system that wouldn't spawn OpenMP threads. The compiler was fine. The binary would run ok in another PC. But when I tried to run it on that specific PC, it wouldn't multi-thread at all.

I was using Intel's ifort compiler, so I don't really know why that happened

[Chaz]

Yeap. The original vote was kind of an attempt to get things rolling ? if they vote on the system to choose, they can then start thinking about further plans of transitioning etc. since they have a point of reference. But now Ian Jackson wants the vote to include several other things at the same time, making things more complicated (talking about something called "facts on the ground", I have no idea what that means). And so people are now arguing about what points the vote should include, and how the points should be worded... And whatever happens, a GR seems to be inevitable anyway.

"Facts on the ground", and especially "creating facts on the ground" is a reference to Israel's policy of evicting Palestinians from land in the West Bank inside Palestine and building Jews-only settlements on the land. Israel then argues that since the land is now inhabited by Israelis, it must become part of Israel. The idea is that with Jewish settlements all over the West Bank it will become impossible in a practical and political sense to ever restore Palestinian control over the entire West Bank. Instead the West Bank would stay under permanent Israeli control or be divided with Israel controlling the best land, the water, and the strategic points and Palestinians confined to small ghettos.

https://en.wikipedia.org/wiki/Facts_on_the_ground

It is used to suggest that one party is acting/will act in bad faith in a way which forces the argument to go their way. Ian is probably afraid that developers will make their projects only support System D (systemd to people who speak Programmer rather than English) as a way to pressure users and maintainers into using System D. He probably has Gnome in mind.

[chrisq]

I remember when ArchLinux had introduced systemd, they lose 30% of users Good Luck

Personally I switched to Arch for systemd, while not the only reason, not having it would have been a deal breaker.

[Delgarde]

Another crazy thing is, that instead of breaking the complex questions down into simple simple questions, resolving one problem at a time like chairman Bdale suggested, they now have an extremely complicated ballot, wide open for different interpretations.

Well, it's not entirely crazy - they don't want to be making one big decision, only to have to revise it once they start looking at answering subsequent questions.

But yes, a lot of the decisions they need to make are the same regardless of whether they choose systemd or upstart, and it would be helpful if they could at resolve those questions to simplify the main issue. As is, they're trying to treat everything as one interconnected big-picture thing, one which I think most of the committee are struggling to get their heads around.

[Sergio]

Why are these the only options? What about, say, launchd (the "original" systemd)?

[Awesomeness]

Systemd will win. upstart was good for long time, not the better solution now.

Canonical already bought some votes. They can buy more.
I doubt than any systemd proponent will buy systemd votes. Many corporate systemd proponents are competing against Debian and adopting Upstart could result in weakening Debian, ie. benefiting the competition.

[Ericg]

Why are these the only options? What about, say, launchd (the "original" systemd)?

Launchd doesn't currently run on Linux?

SysV -- Works but showing its age, could do better. No change
OpenRC -- Only really used by one distro AFAIK, Gentoo, not much different than SysV... If they're gonna change may as well go big or go home.
Upstart -- Only really used by one distro (and its Derivs), some Devs have problems with CLA, big change
Systemd -- Lots of distros using, big change.
Launchd -- Doesn't run on Linux, its JUST starting to get ported to FreeBSD, no ones even working on it. Also Systemd took Launchd and tried to fix its perceived shortcomings, so why use the prototype when the finished product is available?

[justmy2cents]

You're forgetting Debian jessie, for a stable bleeding edge distribution. Never mind that you can also use pinning with Debian apt, to keep your base system as the stable distribution, while cherry-picking packages from testing, unstable, and experimental for installation. Works beautifully, and is a great way to keep your system rock stable, while enjoying the latest versions of whatever you need. That's something Fedora cannot do, from my knowledge.

cherry picking... after promoting stable? @.@' dear... god... that... never... works... out... sigh, you end up with something that is even less stable than any bleeding edge.

about your knowledge

what would be the point of pinning in fedora? whole repo stacks are rebuilt against every new release. if you want testing, simply enable testing repo and those packages will be used. then again, what would be the point if new release is every 6 months or so? packages are never old enough to warrant that.

what you get with fedora is stabilized new things every 6 months. check out how fedora process is done. also, check out how some mindless people bugger off fedora since it is slipping shipping dates and check why it does that. if it would be pure bleeding edge it would simply ship as it is at specified date, slipping dates are solely for the sake of not getting trough QA.

my oldest home server still runs on f13. running 5 virtual machines and all possible external services with 0 downtime in whole this time. yep, unstable as hell. would i run fedora in production? nope, centos|rh all the way there, if nothing else than just more QA. but, if you want everything working out of the box for user box, try Korora which is basically still same Fedora with different starting point.

[Delgarde]

Why are these the only options? What about, say, launchd (the "original" systemd)?

Because adding something like that would just complicate things further. As things stand, the split between Upstart and Systemd is bad enough - on the one hand, the option favoured by Ubuntu, one of Debian's biggest supporters; on the other, the option favoured by just about everyone else. Launchd is out of the question because it's just not a serious contender (much the same as OpenRC, which was briefly looked at, but not deemed worthy of further consideration).

[interested]

Well, it's not entirely crazy - they don't want to be making one big decision, only to have to revise it once they start looking at answering subsequent questions.

But yes, a lot of the decisions they need to make are the same regardless of whether they choose systemd or upstart, and it would be helpful if they could at resolve those questions to simplify the main issue. As is, they're trying to treat everything as one interconnected big-picture thing, one which I think most of the committee are struggling to get their heads around.

Well, now they have to decide whether Linux-Jessie should support multiple init systems, or just one, or if more than one init-system, whether that support should be mandatory or volunteer, or whether there should be a default init-system or not, besides that, they shall also decide whether what exactly what particular init-system to choose in each of the above cases, all on one voting ballot. Ian has also tagged on a "should programs be banned from supporting systemd" (the pid 1 clause) if there is more than one init system.

If they had broken down the questions into smaller portions, they could have much easier ballots like:
1. Which init system as default for Linux-Jessie
2. Should Linux Jessie support more than one init system or not
3. If more than one init system, should the support be voluntary or mandatory

Then they could turn their attention to the Hurd/kFreebsd ports, and perhaps actually listen to what they want. As the Ian Jackson's ballot is constructed right now, they risk being dictated a init system they don't care for at all.

This link isn't a official statement from the Hurd/kFreebsd porters, but just a draft, but that they mostly prefer OpenRC and really don't want Upstart, has always been my impression too:

http://thread.gmane.org/gmane.linux.debian.ports.hurd/15235

I can't see why they couldn't break down a complex question into smaller easier ones, or why they would ever have to re-vote on any of those smaller issues. In fact, voting for one issue at a time seems to naturally resolve other questions further down the decision tree.

OTOH, Ian's voting suggestion is perfect for splitting votes, and letting sub-questions ride along the main question; eg. in order not to loose the vote, all the systemd proponents would have to vote "systemd and mandatory support for multiple init-systems", even though 3 out of 4 really wanted a single default init-system with voluntary support for other systems.