There are a bunch of better alternatives out there.
You can argue that he does that indirectly: He likes that systemd provides something like a service layer, which the tools I mentioned don't really do. Maybe you mean that. That would work, I just happen to think he is wrong there. That's not the job of init, making it a plus point of an init system is some kind of mental gymnastics I don't agree with (though I liked that talk).
The problem with systemd is that it's replaced systems that met peoples needs with a single system that many people consider to be worse.
Perhaps there are some niche environments that systemd is needed, and that's great -- but why break things for everyone else?
> The problem with [old init] is that it's replaced systems that met people's needs [systemd] with a single system [old init] that many people consider to be worse. > > Perhaps there are some niche environments that [old init] is needed, and that's great -- but why break things for everyone else?
So the argument is kind of meaningless without a tie-breaker of some kind; perhaps popularity vote.
I think you're under the impression that your views of systemd's relative suitability are held by the majority of users, but I believe that is incorrect.
If you sell something and people choose to use it to solve a problem, that's great. If you force change on someone, and not only not deliver any benefit to them, but actually make things worse, expect pushback.
Remember too that most people's systemd impact wasn't just replacing init, it was a whole raft of changes from log formats to dns.
Imagine you had a working systemd setup and someone came along and replaced everything. They claim that there are use cases it's better for, but you don't see yourself with those use cases. You would be averse to that change. You then find out that pretty much every distribution going has got rid of systemd. I suspect you wouldn't be happy.