> TL;DR by analogy =
> - X, which like systemd, did eevrything in a giant sphagettified mess. (But 
> still missed out on the sound... and used VTs)
> - wayland (library + compositor) + libinput + pipewire + wireplumber + 
> whatever-else is the future.
> - We needed a desparate solution like X(systemd), but the *real* solution is 
> wayland(split-up components)
> - We need to learn from systemd, but be more "clean" and "agnostic".
> - Initially the systemd replacement will be broken just like wayland was and 
> X was fine, but it will improve, to be better (not for everyone).
> - That's it. Period. "text-only => X => wayland" is (almost) exactly analogus 
> to "sysVrc => systemd => dinit + 66 + turnstiled + (a new acpid) + seatd + 
> whatever-else"

and openrc?

> The opentmpfiles CVE issue, BTW, could have been fixed by adding a `-h` to 
> all `chown` calls. AND `sysctl` `fs.protected{sym,hard}lniks=1` (but this is 
> default anyways).
> The *real* issue was lack of maintenance, many new features weren't supported.

this is simply not true. `-h` only applies for the final component of a
path, not anything else inbetween. there's no way to prevent the cve
with posix shell tools alone, c code *is* necessary.

> The rest of systemd:
> - `systemd-boot` (maybe `systemd-ukify` too) and `systemd-udevd` are pretty 
> much the *only* pieces of systemd which have any requirement without a 
> replacement needed.

neither really requires replacement at the moment, there's no issues
with them except "but they come from the systemd tree!".

> - `systemd-journald` is a nice concept, but poorly implemented. Re-implement 
> is properly or throw it out of the window.

let's please not do binary logs again.

> Missing pieces:
> - CGroups, well... can be implemented using a helper command...

most if not all cgroup control can be done directly by shell, or c
filesystem functions, what you need is integration. openrc has some for
it's services although it needs improvements.

> - `openrc-settingsd` needs to follow the "standard" files... and needs a CLI.

feel free to list what "standard" files it doesn't follow, i can
implement that.

> - `turnstiled` is yet to have a session-info-query library, and a CLI

turnstile is really incomplete at the moment, i doubt it'd get anywhere,
and i'm somewhat unsure about it's direction.

> - `66-dbus-launch` is for `66`, a replacement for `dbus-broker-launcher`. A 
> similar but service-manager-agnostic thing (or service-manager-specific 
> things)

dbus itself needs better launcher integration, dbus-broker is linux
specific and most users don't run it.

Attachment: 68990292A7A98C5E.asc
Description: application/pgp-keys

Attachment: signature.asc
Description: PGP signature

Reply via email to