Debian Still Debating Systemd vs. Upstart Init System

[Ibidem]

I hope that a choice is made for technical reasons (i.e. that a given solution best suits Debian's technical needs), and not for political reasons (such as to attempt to keep another solution from becoming dominant). I'm not suggesting that the layman, onlooking discussion surrounding this bears any indication of the human natures of the members of the technical committee, but much of the layman discussion going on various places has been political. I also hope that a one-size-fits-all solution isn't attempted to be hammered into place, unless one size truly best fits all. If the maintenance cost can be borne, I hope they choose what is best for Debian GNU/Linux, Debian GNU/kFreeBSD, and Debian GNU/Hurd, considered individually, rather than the lowest common denominator. For the Debian Project, Linux is by far the most used of the three, followed by kFreeBSD, and then Hurd.

The Debian Project is not the community arm of any company, and I think any committee member of Debian Project should recuse themselves as a matter of clearly defined, predetermined policy when tight conflict of interest arises. This doesn't mean that committee members of the Debian Project are any more conflicted or less professional than others, it's just good form for any independent organization. When a conflict of interest is so apparent that some organizational members feel obligated to give the typical 'we have a handle on it' - 'we rise above conflict of interest' speech, that's a red flag.

I suggest you read https://lists.debian.org/debian-deve.../msg01014.html and following as well.
For those not familiar with the people involved, the poster (Ian Jackson) is the "former Canonical developer"; he wrote the Debian constitution back before upstart was thought of; the fellow he's quoting is Debian's systemd maintainer, the freedesktop.org sysadmin, and a systemd developer.

Anyhow: the work systemd avoids on Linux would still have to be done before a release, since kfreebsd is an official port.

[eidolon]

I suggest you read https://lists.debian.org/debian-deve.../msg01014.html and following as well.

I already did, and had taken it into consideration, and am still unpersuaded that it is for the best, whether with regard to the Debian Project, or other organizations where similar arguments would apply.

[valeriodean]

Nope, only if they want to upstream those changes.

In fact they already maintain some patches downstream because the authors don't want to sign the canonical CLA.
To me, this is another point against upstart.
Not only they continuously take idea and work from the systemd team (as per logind) but also if Debian takes upstart, they need to maintain patches downstream instead of send it upstream, because the CLA.
So not only the ugly combination upstart+logind (<= 204), but also the patches downstream? Oh, came on.

To despite of all, I think that the TC will go with upstart.
Meantime, I'm sure, that without 3 people in the TC with a past of upstart developer and canonical's employees, the TC would decide for systemd.

[vbtux]

I suggest you read https://lists.debian.org/debian-deve.../msg01014.html and following as well.
For those not familiar with the people involved, the poster (Ian Jackson) is the "former Canonical developer"; he wrote the Debian constitution back before upstart was thought of; the fellow he's quoting is Debian's systemd maintainer, the freedesktop.org sysadmin, and a systemd developer.

Anyhow: the work systemd avoids on Linux would still have to be done before a release, since kfreebsd is an official port.

No, the release team will decide if kfreebsd is an official port for Jessie the 28th of january! Same for other architectures.
Note that the release team is concerned by the high amount of FTBFS generated by kfreebsd.

Also, Russ (ctte member in favor of systemd) has proposed to have upstart on kfreebsd and systemd on Linux.

[vbtux]

I suggest you read https://lists.debian.org/debian-deve.../msg01014.html and following as well.
For those not familiar with the people involved, the poster (Ian Jackson) is the "former Canonical developer"; he wrote the Debian constitution back before upstart was thought of; the fellow he's quoting is Debian's systemd maintainer, the freedesktop.org sysadmin, and a systemd developer.

Anyhow: the work systemd avoids on Linux would still have to be done before a release, since kfreebsd is an official port.

No, the release team will decide if kfreebsd is an official port for Jessie the 28th of january! Same for other architectures.
Note that the release team is concerned by the high amount of FTBFS generated by kfreebsd.

Also, Russ (ctte member in favor of systemd) has proposed to have upstart on kfreebsd and systemd on Linux, simply because it'd be a shame to not exploit systemd features due to its non-portability!
systemd has so much advantages compared to upstart, that this non-portability has not enough weight.

The question that we should ask ourselves (Debian developers) is : Should we take the risk to not improve our Linux port because a port with less than 0.9% of Debian users can't gain systemd support?

[GreatEmerald]

I already did, and had taken it into consideration, and am still unpersuaded that it is for the best, whether with regard to the Debian Project, or other organizations where similar arguments would apply.

Yea, same here. If they were just Upstart contributors, it would be fair point. But if they're also Canonical employees, then it's a no-go. Canonical desperately needs Upstart to gain more popularity, and with everyone siding with systemd and only Debian being undecided, this opportunity is very important for Canonical. Plus, since they are Canonical employees, the rationale that "they work on it because they think it's good" doesn't hold ? in a company you are told what to do, and you do it. And once you do it, you're familiar with it and thus you vote for it. I don't doubt that those people sincerely think that Upstart is the better choice, but they were convinced about its superiority through having to work on it.

On the other hand, Russ Allbery, as he nicely showed, is completely unaffiliated with either, but he did active research on both and his decision is clearly technical.

[vbtux]

Yea, same here. If they were just Upstart contributors, it would be fair point. But if they're also Canonical employees, then it's a no-go. Canonical desperately needs Upstart to gain more popularity, and with everyone siding with systemd and only Debian being undecided, this opportunity is very important for Canonical. Plus, since they are Canonical employees, the rationale that "they work on it because they think it's good" doesn't hold ? in a company you are told what to do, and you do it. And once you do it, you're familiar with it and thus you vote for it. I don't doubt that those people sincerely think that Upstart is the better choice, but they were convinced about its superiority through having to work on it.

On the other hand, Russ Allbery, as he nicely showed, is completely unaffiliated with either, but he did active research on both and his decision is clearly technical.

Absolutely, Russ and Colin (Canonical employee) have made impressive research, contrary to Ian Jackson who mostly bring fallacies. But we are used to his behavior towards GNOME, Network Manager, systemd,
and every other projects that Lennart touch (see older tech commitee discussions).
You should take a look at the debian-ctte mailing list, the guy already want to split the systemd package despite that other tech ctte members have not expressed their wish! He acts like a despot.

[GreatEmerald]

Absolutely, Russ and Colin (Canonical employee) have made impressive research, contrary to Ian Jackson who mostly bring fallacies. But we are used to his behavior towards GNOME, Network Manager, systemd,
and every other projects that Lennart touch (see older tech commitee discussions).
You should take a look at the debian-ctte mailing list, the guy already want to split the systemd package despite that other tech ctte members have not expressed their wish! He acts like a despot.

Well, Colin Watson doesn't make a convincing case for me, either. He stated multiple times that he is against systemd because he's against tight coupling ("But I think if I have a pre-existing bias it is more towards not having a monoculture"), and that's not a technical reason and falls squarely into the "Debian should send a message" category. I also agree with Uoti Urpala, in that his in his overview the two example issues he outlined are not issues whatsoever.

And I also find the fact that he sees the Upstart event model as the best to be further proof of what I said before ? this is more of a vote of familiarity, not a technical decision. Debating the two models is like debating low-level vs high-level programming languages. The systemd approach is way easier to understand (again I agree with Uoti Urpala, "Start Y, and when Y is ready, start X" is not how people understand hierarchy; this would raise the question "Why don't you just start X?" and the answer would be either the same "it needs task Y", or "because that's how it should be" ? which is a meaningless answer to humans). The Upstart model may be more flexible, but it's also more low-level and thus hard to understand for regular users.

[vbtux]

Well, Colin Watson doesn't make a convincing case for me, either. He stated multiple times that he is against systemd because he's against tight coupling ("But I think if I have a pre-existing bias it is more towards not having a monoculture"), and that's not a technical reason and falls squarely into the "Debian should send a message" category. I also agree with Uoti Urpala, in that his in his overview the two example issues he outlined are not issues whatsoever.

And I also find the fact that he sees the Upstart event model as the best to be further proof of what I said before ? this is more of a vote of familiarity, not a technical decision. Debating the two models is like debating low-level vs high-level programming languages. The systemd approach is way easier to understand (again I agree with Uoti Urpala, "Start Y, and when Y is ready, start X" is not how people understand hierarchy; this would raise the question "Why don't you just start X?" and the answer would be either the same "it needs task Y", or "because that's how it should be" ? which is a meaningless answer to humans). The Upstart model may be more flexible, but it's also more low-level and thus hard to understand for regular users.

Yeah I may be wrong, but still, I have the feeling that he is more sincere than Ian Jackson and Steve Langasek who play dirty political game!

[vbtux]

Last bit from Ian Jackson:

It seems unlikely
that there would be a majority in the TC for picking systemd, and
separately a majority in the TC for overruling the systemd maintainers'
refusal to implement a simpler readiness protocol.

So a decision to pick systemd automatically comes with the expectation
that all daemons will get the new build- and runtime dependencies, and
maintainers will be expected to accept those patches.

Isn't it shameful? There are still 4 TC members (Keith Parckard, Bdale Garbee, Don Armstrong, Andreas Barth) that must share their choice, but the guy is already saying that
systemd will not not be picked.