Services that can start by default policy feedback

Matthew Garrett mjg59 at srcf.ucam.org
Fri Feb 25 18:22:25 UTC 2011


On Fri, Feb 25, 2011 at 09:46:08AM -0800, Toshio Kuratomi wrote:
> On Thu, Feb 24, 2011 at 06:32:44PM +0000, Matthew Garrett wrote:
> > There are no essential services, which means any proposal that contains 
> > the phrase "non-essential services" is already unimplementable.
> > 
> You've said this many times and it seems that you do it to be
> obstructionist.  The constructive way to deal with this is to start making
> a list of what people really mean by "essential" and then propose alternate
> words to use.

Like Jesse said, my objection here is that using the word "essential" 
just results in us being doomed to argue over what "essential" means. 
A literal interpretation of "essential" means "start init and have it 
launch a getty". I don't think anyone's advocating that that be the 
outcome of a vanilla Fedora install. An alternative would be "Essential 
for a traditional UNIX experience", which would seem to preclude dbus. I 
don't think that's a rational outcome either. So we end up with 
"Essential for providing an experience consistent with what we feel a 
vanilla Fedora install should provide", which means you haven't actually 
defined "essential" at all. So don't say "essential". Say what you mean.

> I think, by essential, some people mean:
> 
> start the bare minimum so I don't have to start any additional services to:
> 
> ... I don't want anything but init and a shell [*]
> ... log into a getty
> ... log in over the network
> ... log into a desktop
> ... do any client-side operations

That's my point. If people have different interpretations of "essential" 
then any policy using the word "essential" is meaningless. You need to 
define "essential" - and if you're doing that then you don't need to use 
the word in the first place.

-- 
Matthew Garrett | mjg59 at srcf.ucam.org


More information about the devel mailing list