On Tue, Oct 16, 2018 at 05:28:04PM -0400, Gene Heskett wrote:
[...]
> The point is that no one from the systemd camp has, to my knowledge,
> posted a list of things it can do better than sysvinit. Point out
> sysvinit's warts, and what systemd does to excise the ugly stuff.
Now this is unfair. There *has* been a lot of discussion, by both
sides. Lennart Poettering himself (you /may/ like his ways or you
may dislike them) has gone out of his way to explain how the whole
thing works.
Off the top of may head, what systemd has to offer over (out-of-the-box)
sysvinit:
- process supervision: when a daemon is started by sysvinit,
it (typically) plonks its PID in some pid file. sysvinit
forgets about it. If the proc dies, it's up to you to do
something about it (there are "third-party" process supervisors
like runit, or you can just bricolage your own, a thing I've
done myself some times).
This actually brings something back we (those old enough who
weren't wasting their time back then with Norton Commander ;-)
used to have with /etc/rc style booting. Those still showing
some Pavlovian kind of salivation at the sight of "getty"
and "respawn" will know what I mean
- socket-based activation: systemd listens on IPC or network
sockets, and whenever a "client" knocks on them, it starts
the service supposed to be serving this socket: i.e. whenever
someone tries to contact the PosrgreSQL database at port
5432, the database is started.
This takes a lot of manual work usually done to sort the
starting order of services, because they "sort themselves",
based on who needs whom.
This will remind some of us of (x)inetd. It is similar, but
not the same
- Init configuration is expressed in a more declarative way
("what" you want to achieve) instead of the traditional
imperative way ("how" you get there), which is deemed to
be more robust wrt. change.
Now there are counter-arguments to all of that, and on my balance
they dominate (that's why I stick with sysvinit), but saying that
"no one from the systemd camp has ... posted a list of things ..."
is highly unfair. Look again.
You ask where? My go-to site for fundamental knowledge is lwn.net.
A (hpoefully *not* Google) network search "systemd site:lwn.net"
yields, among other things, those gems:
Systemd programming part 1: modularity and configuration
https://lwn.net/Articles/584175/
Systemd programming part 2: activation and language issues
https://lwn.net/Articles/587385/
Enjoy. And apologize to the systemd proponents ;-)
(And all of this from a systemd opponent. Sheesh.)
Cheers
-- tomás
Attachment:
signature.asc
Description: Digital signature