[@core] working definition for the minimal package set

[Steve Grubb]

On Tuesday, November 13, 2012 04:55:50 PM Adam Williamson wrote:
> > So far everything works without, and I think we should endevor to keep
> > that true.
> 
> I think this is similar to the firewalld issue in that the basic theory
> here is that, look, NetworkManager is the way, the truth and the light:
> it's supposed to be the One True Networking System, and we're just
> keeping the network service around because there's some stuff it does
> that NM doesn't do yet.
> 
> This logic is getting a tad stretched because we've been rolling with it
> for several years at this point, but AIUI this is still the party line
> and the reason NetworkManager is in core. In theory the idea is not that
> we provide, actively maintain and support both NM and the network
> service, but that we want to only provide, maintain and support NM, and
> we're keeping the legacy 'network service' stuff around only until NM is
> done.
> 
> It might be worth re-evaluating whether that's realistic any more,
> though, and whether we're really committed to finally replacing
> network with NM in some kind of reasonable timeframe.

For Common Criteria purposes, everything running as root goes under the 
microscope and its painful and costly. We have to avoid that. If NM did not 
run as root and just retained whatever capability it needed, then we have an 
easier time. Same thing with firewalld.

-Steve