The deep underlying problem is that the Linux community has a culture that is set by people like Linus, Andy Tridgell, Hans Reiser, Alan Cox, Miguel de Icaza and so on. And they live in an ecosystem with more sociopaths like Larry McVoy, Eric Raymond, RMS, and so on. Many of them are really fine software coders, and smart people who can keep enormous complexity in their head, and have an overview over huge systems. But the basic problem is that these people have little understanding of computer use in the real world, hugely inflated opinions of their own importance, and no system of checks and balances that can tell them "you're wrong". Two words: arrogant narcissists. That's one of the reasons why smallish disagreements (like the infamous source control battle between Linus/Andy/Larry over Bitkeeper versus git) can not be resolved sensibly, and turn into personal feuds of an intensity not seen since the middle ages.
There are some perfectly reasonable and nice people in leadership positions of the Gnu/Linux/OSS ecosystem too: Ted Ts'o and Stephen Tweedie come to mind, although I haven't seen Stephen in years; Guido van Rossum also has a reputation of being a good and sensible person (haven't met him though). Valerie Aurora (formerly Henson, of file system fame) has written some insightful pieces on the destructive culture of the Linux community, with particular emphasis of its rather shoddy treatment of women. This has led to her leaving the development of Linux.
Yet, this is the software we are stuck with. Just using systemd as an example: whether we like it or not, whether it has way too many bugs, and whether Lennart is an idiot with good typing skills, we'll use it, because the ecosystem has no other solution. Example: I have an RPi3 at home, and a few months ago I gave up on running FreeBSD on it, because too many things were too difficult, things that just run seamlessly under Raspbian. So now I'm stuck with Linux, and as of the week before Christmas my Pi is actually in "production" (installed in my pump shed, monitoring water tank level and pump pressure). But I got tired of having to restart the monitoring program after every power outage by hand, so I wrote a service description file for systemd. Took me over half an hour to get right, but fortunately the web is full of pages (mostly half true) that explain how to do it. Now it works. On FreeBSD it would have taken one minute, but that's only because there I have the daily experience of administering the system. The important part is: in many scenarios (like wanting to run on an RPi3 with wireless) there is no alternative to Linux, and we'll use systemd. And with a little effort and some gritting of teeth, systemd is usable.
Which has no bearing on whether FreeBSD should use systemd or not. My personal opinion is that it should not, because otherwise it will just become a Linux clone with a slightly different kernel implementation. Instead, FreeBSD could tune and improve the existing init system to match some of the benefits (like defaulting to parallel init for faster boot times). This may lead to incompatibilities with desktop environments and end-user tools that rely on systemd, but maybe people who need/want those things will be better off in the Linux ecosystem.
There are some perfectly reasonable and nice people in leadership positions of the Gnu/Linux/OSS ecosystem too: Ted Ts'o and Stephen Tweedie come to mind, although I haven't seen Stephen in years; Guido van Rossum also has a reputation of being a good and sensible person (haven't met him though). Valerie Aurora (formerly Henson, of file system fame) has written some insightful pieces on the destructive culture of the Linux community, with particular emphasis of its rather shoddy treatment of women. This has led to her leaving the development of Linux.
Yet, this is the software we are stuck with. Just using systemd as an example: whether we like it or not, whether it has way too many bugs, and whether Lennart is an idiot with good typing skills, we'll use it, because the ecosystem has no other solution. Example: I have an RPi3 at home, and a few months ago I gave up on running FreeBSD on it, because too many things were too difficult, things that just run seamlessly under Raspbian. So now I'm stuck with Linux, and as of the week before Christmas my Pi is actually in "production" (installed in my pump shed, monitoring water tank level and pump pressure). But I got tired of having to restart the monitoring program after every power outage by hand, so I wrote a service description file for systemd. Took me over half an hour to get right, but fortunately the web is full of pages (mostly half true) that explain how to do it. Now it works. On FreeBSD it would have taken one minute, but that's only because there I have the daily experience of administering the system. The important part is: in many scenarios (like wanting to run on an RPi3 with wireless) there is no alternative to Linux, and we'll use systemd. And with a little effort and some gritting of teeth, systemd is usable.
Which has no bearing on whether FreeBSD should use systemd or not. My personal opinion is that it should not, because otherwise it will just become a Linux clone with a slightly different kernel implementation. Instead, FreeBSD could tune and improve the existing init system to match some of the benefits (like defaulting to parallel init for faster boot times). This may lead to incompatibilities with desktop environments and end-user tools that rely on systemd, but maybe people who need/want those things will be better off in the Linux ecosystem.
) can start up in parallel. So you would have a series of serial and parallel startups. It would be more reliable I think than systemd ever could be. As an added benefit, it wouldn't be hard to make the modifications to make it work that way either, and there is no inherent race condition, so the behavior is predictable.
