This whole debate saddens me

Posted Dec 2, 2014 12:56 UTC (Tue) by mchapman (subscriber, #66589)
In reply to: This whole debate saddens me by nix
Parent article: The "Devuan" Debian fork

> So... almost all this stupid, tiresome systemd arguing could have been avoided if the systemd people had created a shared library (like anyone else would have) and set up more than one git repo?

Possibly, possibly not. It might be worthwhile considering what the trade-offs there are. Distributions are already shipping networkd and resolved in separate packages from systemd-the-init, so there has to be a compelling reason why you'd add an additional shared library for these components to use.

One advantage to having everything in the one Git tree is that code can be moved in and out of the shared directory at will. If you have things in separate trees then this becomes a lot more difficult -- suddenly you have to *version* your internal APIs, not just your external ones.

Also, just in case you're wondering, while there is some overhead in having multiple binaries linking the same set of shared source files, it isn't too much. systemd is compiled with function section (or with link-time optimisation), which means unused code is completely dropped. Each binary gets only the shared functions it actually *uses*.

This whole debate saddens me

Posted Dec 2, 2014 17:00 UTC (Tue) by johannbg (guest, #65743) [Link] (4 responses)

"Distributions are already shipping networkd and resolved in separate packages from systemd-the-init"

Which distribution are that?

This whole debate saddens me

Posted Dec 3, 2014 0:17 UTC (Wed) by mchapman (subscriber, #66589) [Link] (3 responses)

> Which distribution are that?

Ah, you got me.

I was getting confused. I actually had a particular Fedora COPR repository for RHEL 7 in mind: https://copr.fedoraproject.org/coprs/lnykryn/systemd/ . It's currently got networkd in a subpackage. resolved is currently disabled, though as far as I can see there shouldn't be anything stopping it from being put in a subpackage.

This whole debate saddens me

Posted Dec 3, 2014 13:56 UTC (Wed) by johannbg (guest, #65743) [Link] (2 responses)

Zbyszek decides how he indents to maintain systemd in Fedora until someone escalates the packagesplit ( which at is point,after all this time would be a bit silly from my pov ) to FESCo, which in turn would most likely request input from BaseWG ( which probably would go for the smallest core/base footprint which means not split it ) followed by FESCo rule based upon that and his maintainership being overruled and then each WG overule the overule and decide for themselves if they want to split it ( or not ).

Lukáš RHEL 7 and derivatives corp repo that you refer to is only intended to be testing repository of new systemd releases ( I think he's already drowning in backporting patches for the release being shipped in RHEL 7 with RHEL 7 being bound to be stuck on that release for the next 10 years or so ).

This whole debate saddens me

Posted Dec 3, 2014 23:48 UTC (Wed) by mchapman (subscriber, #66589) [Link] (1 responses)

Sure, but that's all kind of irrelevant to this thread.

The context of my earlier post was about the modularisation of systemd, and whether having a shared library for its internal APIs (rather than a source library) would be worth it. The existence of this repository indicates that a shared library isn't even necessary to package systemd's components separately.

This whole debate saddens me

Posted Dec 4, 2014 0:39 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]

True. Somebody who is interested in it could talk to Fedora maintainer for example and submit a spec file patch. It shouldn't too hard.