Interview: Lennart Poettering

[Randicus Draco Albus]

It does not need to be about taking control. At least in theory.

Red Hat want to create and adopt a new core system; they can either modify every piece of third-party software for their distribution or third-party developers can make two versions - one for Red Hat and one for everyone else; the latter is easier for them.

So convince every other major distribution to adopt it for their own benefit? That would not be a takeover conspiracy. It would simply be a company trying to make things easier for themselves, but would result in the same thing. Myself, I doubt the development team's eagerness to force systemd on the "haters" is theirs alone. Why would their employer want them to waste effort trying to force everyone else to use it, instead of focusing only on what is best for the employer? They may not desire to take over everything, but something is afoot. At least that is the way I see it.

[cynwulf]

It does not need to be about taking control. At least in theory.

Red Hat want to create and adopt a new core system; they can either modify every piece of third-party software for their distribution or third-party developers can make two versions - one for Red Hat and one for everyone else; the latter is easier for them.

Developers will support systemd and if they want to, they can support sysvinit, openrc or anything else. In fact there's not much that developers need to change as little is tied to the init system. The developers building in support for the other systemd stuff, such as the session management stuff are not necessarily breaking the old way of doing session management. For your service to run on systemd, as far as I know, you only need to write unit file and a few other bits to start it.

So convince every other major distribution to adopt it for their own benefit? That would not be a takeover conspiracy. It would simply be a company trying to make things easier for themselves, but would result in the same thing. Myself, I doubt the development team's eagerness to force systemd on the "haters" is theirs alone. Why would their employer want them to waste effort trying to force everyone else to use it, instead of focusing only on what is best for the employer? They may not desire to take over everything, but something is afoot. At least that is the way I see it.

The thing is that Red Hat haven't even been pushing it. You make it seem like men in red fedoras carrying brief cases have been out bribing Debian, SUSE and Arch developers... Red Hat have not told anyone to use it. Arch adopted it because Arch will go for the latest bleeding edge software no matter what. butnut were essentially already developing and using the same kind of thing since '06 - namely upstart - once Debian adopted systemd, butnut just went with that as it's simpler for them to maintain.

We have to face facts that Debian has been heavily influenced by butnut people for several years now and it was a butnut agenda which was behind pushing upstart OR systemd into Debian, not Red Hat. Because so many Linux distros are based on Debian/butnut, that effectively means instant and massive systemd adoption. Before that it was mainly just Arch (one of the earliest adopters), fedora (obviously), SUSE and only one or two others.

So I can't see where the developers are 'forcing' any one to do anything? Torvalds is not going to make the kernel only function with systemd - not going to happen. The big desktops are aimed at the kind of users who will probably accept systemd anyway (there is little point in running a beast like gnome and moaning about the un-UNIX-like nature of systemd is there?), those developers are already having to maintain equivalent functionality, so will probably welcome offloading that onto someone else.

When all is said and done, some distros adopted this and some didn't. Most of the FUD is coming from the hysterical anti systemd camp - these are people who thrive on creating conspiracies and leading other people on, the other forum was full of this kind of posting recently. Another large part of the problem is the lead developer's attitude - but some people take his comments far too seriously.

"init choice" is also an idiotic argument. Architectures have been dropped in the past and there was less of a hullabaloo. Personally I had no interest in choosing init - I just wanted the established working, stable and proven init system. I didn't mind systemd being in the repos and others being able to install it, I didn't mind if the odd lib got pulled in so long as my init system wasn't switched over.

Some people object to that - in fact some of the anti systemd people seem to object to it being developed at all - they essentially want them to stop what they're doing. The thing about free software is that as with freedom of speech, people need to be free to develop stuff you don't like as well.

If these people just voted with their feet and moved to and supported the distributions that are committed to avoiding systemd or preserving choice, well that would be something...

[mean_dean]

love the avatar by the way

edit - I did NOT mean that literally

[Randicus Draco Albus]

"init choice" is also an idiotic argument.

This is the key factor for me. I refuse to believe systemd was originally intended to only be an init process and grew of its own accord. It is a great stretch to believe systemd was not intended to be all-encompassing from the beginning. If some consider that a conspiracy theory, so be it, but I find it difficult to believe that a simple init project took on a life of its own and hungered to control the rest of the system. But then, perhaps I am conspiracy-minded.

[Randicus Draco Albus]

love the avatar by the way

I wonder how got that photograph of me!

[debil]

Perhaps a question Poettering should be asked in a future interview is why systemd was not developed from start to finish before being released, instead of being implemented on systems piecemeal while it is still being developed?

Well, the answer is pretty guessable since it's about a well-known software development philosophy.

[Randicus Draco Albus]

Such a philosophy might work with small projects - might -, but I question its feasibility with large projects.

[debil]

Such a philosophy might work with small projects - might -, but I question its feasibility with large projects.

It's also the development philosophy behind Linux kernel, for example. Isn't it a project large enough?

[cynwulf]

It's probably worth mentioning that OpenBSD adheres to this philosophy... and releases every 6 months to deadline.

[Randicus Draco Albus]

It's probably worth mentioning that OpenBSD adheres to this philosophy... and releases every 6 months to deadline.

But how many changes do they make? They are not redesigning the entire OS. To me it is more analogous to tweaking than releasing. But I could be wrong.

It's also the development philosophy behind Linux kernel, for example. Isn't it a project large enough?

One of the biggest criticisms non-Linux people have of Linux is that the kernel is a "large mess of code that works due to luck more than by design." I do not know enough about the kernel to know if that criticism is accurate, but that is what people say.

The point is I am not a fan of the philosophy. I believe it is better to take a little extra time to release a finished product. I have a conservative attitude toward such things.