F24 System Wide Change: Systemd package split

Zbigniew Jędrzejewski-Szmek zbyszek at in.waw.pl
Fri Nov 20 02:18:17 UTC 2015


On Thu, Nov 19, 2015 at 02:51:39PM -0800, Andrew Lutomirski wrote:
> On Thu, Nov 19, 2015 at 7:28 AM, Zbigniew Jędrzejewski-Szmek
> <zbyszek at in.waw.pl> wrote:
> > On Thu, Nov 19, 2015 at 03:40:53PM +0100, Kevin Kofler wrote:
> >> Zbigniew Jędrzejewski-Szmek wrote:
> >> > networkd+resolved is about 1MB, and more importantly, they do not
> >> > bring in extra dependencies. We discussed also splitting that out on
> >> > the upstream mailing list, but in the end the gain didn't seem important
> >> > enough.
> >>
> >> From a cleanliness standpoint, it still makes sense to split it out. Also
> >> because this is network-facing code and thus potentially security-relevant.
> >
> > It's not network facing code. It configures the network, but exits
> > after applying the configuration, and does not listen for incoming packets.
> > Also, unless you actually provide some configuration (*.network,
> > *.netdev files), it doesn't do anything.
> >
> > I need something more convincing than general "cleanliness". systemd
> > has many many binaries, and splitting each out into a seperate package
> > without some noticable gain would be madness.


On Thu, Nov 19, 2015 at 02:51:39PM -0800, Andrew Lutomirski wrote:
> (Also, how is resolved not network-facing?)

You'll notice that I was replying to Neal's question and talking about
networkd, I think it's pretty clear when I say "it configures the network".

> Is resolved considered production-ready code?

No, it's not fully implemented and nobody is recommending it for
general use. OTOH, it doesn't have currently known bugs. But that is
all moot: it is not enabled by default and is not started unless
explicitly configured. It is provided so early adopters can give
it a spin.

Zbyszek


More information about the devel mailing list