It's Back To Voting For The Debian Init System

[justmy2cents]

To my knowledge, all Fedora spins use systemd as the system init..

that was never in question. only session managed DE was talked about

[GreatEmerald]

Since the writing was on the wall, it is hardly surprising that Steve have been back-pedalling on the issue in order to gain maximum influence. His goal is the same; to maximize the Debian support for packages to work without using systemd. That may be because of his genuine belief that this is good for Debian, but it sure is also something that benefit Canonical.

I doubt that. He already quite clearly stated his intent on the matter if Upstart loses the vote (this was even before the individual position statements started rolling in): he will try and support systemd as much as possible, because that allows Debian to have a say in the development of systemd (otherwise, if he tried to make the transition as difficult as possible, systemd upstream would not take Debian seriously). And at the same time he wants to have as much space as feasible to keep maintaining Upstart as long as Canonical stays committed to it (which may not be very long now). So yes, he wants maximum influence, but it really is good for Debian in this case (and doesn't do anything for Canonical either way).

[justmy2cents]

I doubt that. He already quite clearly stated his intent on the matter if Upstart loses the vote (this was even before the individual position statements started rolling in): he will try and support systemd as much as possible, because that allows Debian to have a say in the development of systemd (otherwise, if he tried to make the transition as difficult as possible, systemd upstream would not take Debian seriously). And at the same time he wants to have as much space as feasible to keep maintaining Upstart as long as Canonical stays committed to it (which may not be very long now). So yes, he wants maximum influence, but it really is good for Debian in this case (and doesn't do anything for Canonical either way).

well, without debian, there is no future. according to their presentation of upstart on debconf, there are 4 major users. which looks fine and dandy until you check out the sad truth of that statement
- Ubuntu ... ok, that one was a given
- OpenSuse ... switched to systemd as default at 12.1, beside the fact that upstart was introduced in 11.3 and never as default
- RHEL, Fedora... Fedora moved to upstart as default in 9-15, where it was replaced by systemd and RHEL6 is only RHEL with upstart since 7 is systemd
- ChromeOS... as much as it looks like "here is a big player", once you look at the fact that they still use upstart 1.2 https://chromium.googlesource.com/ch...art-1.2.ebuild they are not even worth considering as being counted.

with debian they would get community adoption and bigger island (aka. being worth to support it), without debian all upstart incompatibility falls on canonical. the differences and cost of patching all projects that will depend on incompatible features will simply be too large to be worth continuing with the project.

debian on the other hand is in really difficult position. multiple kernels. satisfy everyone equally and you hurt largest amount of ppl by delivering "not as good as everyone else". debian linux is simply too large to be put on equal footing as debian bsd/hurd. satisfy majority and hurt minority. in the end it is a decision which can't ever have "good for everyone" solution. kinda "between rock and hard place". pick less than best for linux and watch your linux community starting to vanish elsewhere. pick best for linux and you're bound to see "not fair" arguments

but, more i think off... this is just prelude to next soap opera when debian will have to chose X successor, Mir or Wayland. i'm willing to bet we will see exact duplication of reasoning as we see now from same people

[GreatEmerald]

well, without debian, there is no future.

Which is why Langasek's position makes sense. Might as well jump ship now, before it's too late to make a difference.

debian on the other hand is in really difficult position. multiple kernels. satisfy everyone equally and you hurt largest amount of ppl by delivering "not as good as everyone else". debian linux is simply too large to be put on equal footing as debian bsd/hurd. satisfy majority and hurt minority. in the end it is a decision which can't ever have "good for everyone" solution. kinda "between rock and hard place". pick less than best for linux and watch your linux community starting to vanish elsewhere. pick best for linux and you're bound to see "not fair" arguments

I'm not convinced it's that bad for ports. They get more work to do, yes, but at the same time they get the freedom to decide for themselves, possibly use port-specific inits that work better on them than anything portable would. Besides, if package upstreams don't support the port's init of choice, then you have to question whether those packages were adequately tested on the port to begin with.

but, more i think off... this is just prelude to next soap opera when debian will have to chose X successor, Mir or Wayland. i'm willing to bet we will see exact duplication of reasoning as we see now from same people

That is definitely likely. Ubuntu certainly puts Debian in a rather uncomfortable situation.

[interested]

I doubt that. He already quite clearly stated his intent on the matter if Upstart loses the vote (this was even before the individual position statements started rolling in): he will try and support systemd as much as possible, because that allows Debian to have a say in the development of systemd (otherwise, if he tried to make the transition as difficult as possible, systemd upstream would not take Debian seriously). And at the same time he wants to have as much space as feasible to keep maintaining Upstart as long as Canonical stays committed to it (which may not be very long now). So yes, he wants maximum influence, but it really is good for Debian in this case (and doesn't do anything for Canonical either way).

I don't think the above represents Steve Langasek's opinion. Take for example this quote, which he wrote as a rider to the voting ballot:

"Where would this ballot option rank vis-?-vis FD, for those TC members who
are opposed to the "loose coupling" option?

== dependencies rider version S (split-the-init) ==

This decision is limited to selecting a default initsystem; we
continue to welcome contributions of support for all init systems.

Software outside of an init system's implementation may not require a
specific init system to be pid 1, although degraded operation is
tolerable. Software not part of an init system's implementation may
require interfaces unrelated to service management that are provided as
part of an init system, but the dependency on such interfaces must be
declared in a way that allows the dependency to be satisfied by
compatible implementations on other init systems.

Maintainers are encouraged to accept technically sound patches
to enable improved interoperation with various init systems."

https://lists.debian.org/debian-ctte/2014/02/msg00012.html

While it is a softer version of Ian's L-coupling, it still forces all Debian developers to ensure that their packages will run on non-systemd init systems, even if it means degraded operation.

This may be Steve Langasek's honest opinion, but it is also an opinion that would fit Canonical greatly, since it would mean that all Debian packages would remain compatible with Ubuntu's Upstart. All the patches that the DD's would have to develop to neuter Gnome/KDE/etc. to work on non-systemd systems, would work for Ubuntu too.

I think the above goes far beyond just wanting to maintain Upstart as an alternative init system for Debian.

We may disagree about his motives (and let me add I respect all the CTTE guys, including Steve and Ian), I think it is hard to deny that his opinions are co-aligned with Canonical's present interests.

[mrugiero]

debian on the other hand is in really difficult position. multiple kernels. satisfy everyone equally and you hurt largest amount of ppl by delivering "not as good as everyone else". debian linux is simply too large to be put on equal footing as debian bsd/hurd. satisfy majority and hurt minority. in the end it is a decision which can't ever have "good for everyone" solution. kinda "between rock and hard place". pick less than best for linux and watch your linux community starting to vanish elsewhere. pick best for linux and you're bound to see "not fair" arguments

Pretty much true, except Hurd is probably going to be able to use systemd in the near-mid term. They are working towards providing the features it needs implemented on the kernel (or well, I'm not sure where a micro kernel will put it, my guess is some kind of privileged user land?). kFreeBSD, on the other hand, would imply Debian themselves implementing what is missing, as FreeBSD upstream is probably not interested in systemd support.

but, more i think off... this is just prelude to next soap opera when debian will have to chose X successor, Mir or Wayland. i'm willing to bet we will see exact duplication of reasoning as we see now from same people

Yes, but you don't have the portability argument here. Both Mir and Wayland require pretty much the same groundwork, so the only portable option (assuming the requirements are not yet accomplished by kFreeBSD and/or Hurd, the latter being more likely to fall behind) is X.org.

[GreatEmerald]

I don't think the above represents Steve Langasek's opinion. Take for example this quote, which he wrote as a rider to the voting ballot:

That is an old quote. You should rather look at his latest proposals. For instance: https://lists.debian.org/debian-ctte.../msg00308.html It's pretty clear that he wants the integration to go forward now, but if possible keep the other options open as well (which makes sense, since he will have to keep maintaining Upstart regardless, and Canonical wouldn't be pleased if he suddenly did a 180 before they themselves decide to switch away from Upstart).

[mrugiero]

That is an old quote. You should rather look at his latest proposals. For instance: https://lists.debian.org/debian-ctte.../msg00308.html It's pretty clear that he wants the integration to go forward now, but if possible keep the other options open as well (which makes sense, since he will have to keep maintaining Upstart regardless, and Canonical wouldn't be pleased if he suddenly did a 180 before they themselves decide to switch away from Upstart).

But if things are as you depict, then there is a conflict of interest. If he is basing his decision in A) the fact he will still need to maintain Upstart (which isn't Debian's problem) and B) that Canonical wouldn't be pleased (again, not Debian's problem), he is taking into account facts that are not relevant to Debian due to his relationship with Canonical.

[justmy2cents]

I'm not convinced it's that bad for ports. They get more work to do, yes, but at the same time they get the freedom to decide for themselves, possibly use port-specific inits that work better on them than anything portable would. Besides, if package upstreams don't support the port's init of choice, then you have to question whether those packages were adequately tested on the port to begin with.

when you put it like this, yea... nothing but positive. have to admit i didn't look at it this way *shame on me*. one can only wish that port maintainers would be more vocal or progressive in this direction you suggest. if everyone took best of his world, debian would gain, not lose as i predicted. kfreebsd going with bsd init might spare them from threads like this one since it would be more with the roots

http://forums.freebsd.org/viewtopic.php?t=35770

it's downright scary to read the treatment like that. i doubt one would find as much downright insulting hatred if they put xbox, ps3 and wii user in one thread where question would be "which is best/worst console and why"

[GreatEmerald]

But if things are as you depict, then there is a conflict of interest. If he is basing his decision in A) the fact he will still need to maintain Upstart (which isn't Debian's problem) and B) that Canonical wouldn't be pleased (again, not Debian's problem), he is taking into account facts that are not relevant to Debian due to his relationship with Canonical.

It's fairly debatable. In this case I think that doing what you must and doing what is right do not clash. The current language proposed by Langasek and Allbery is overall reasonable and is sincere enough. You really can't expect Langasek to propose language that says that alternative inits must be dropped, because I don't think he sincerely wants that, regardless of his status. So I believe his motivation is both sincere and influenced from outside. And while he may have this certain motivation behind his proposal, it is not a hindrance to Debian (not any more, at least), especially compared to what Jackson keeps proposing.

Speaking of Jackson, we haven't heard from him for a while now. I wonder if he's ragequitting...