Well, that's really the point. The one you're using is one of
the (4? 5?)
other zram implementations. It seems a bit more straightforward than the
systemd one for sure.
The zram-generator is probably more straightforward (with literally
less layers of indirection) than what the zram package provides. I
would assume that a generator is also the more idiomatic (and
efficient) solution as far as systemd is concerned and I wouldn't mind
migrating to that if it looked feature-complete and had decent
documentation. There is no manual page[1], only a slightly confusing
README that hints at simplicity and incompleteness.
I don't see the zram-generator as systemd magic nobody knows about
like you claimed. Sure, it's not implemented as a shell script and I'd
need to read its source code to understand what it does given the
current lack of documentation (which is a shame considering how
systemd documentation is often top notch) but as far as the zram
package goes I eventually run into the same situation with zramctl.
If this thread's initiative converges all implementations towards
zram-generator [2] and moves the project forward, then I'm in.
Currently it's a mere rust experiment to me, and by that I don't want
to downplay it, but it's not mature.
Dridi
[1] like most rust programs I have used so far
[2] shouldn't it be called systemd-zram-generator?