Re: discuss: NM server defaults

[Pavel Simerda <psimerda redhat com>]

> From: "Dan Winship" <danw gnome org>
> On 07/01/2013 04:14 PM, Dan Winship wrote:
> > > danw, any rationale behind the argument for ignore-carrier?
> >
> > Servers, by definition, tend to have fixed IP addresses. Therefore, if
> > you are using DHCP on a server, it's probably for ease of deployment,
> > not because you want dynamism.
>
> Any response to this theory?

Mine was posted already. From my point of view, dynamic behavior is the way to achieve the ease of 
deployment. Static mapping of identification to configuration is handled on the server side.

> So currently I am thinking for 0.9.10:
>
>   - flip the default value of monitor-connection-files from true
>     to false for all users, not just server (with a release note)

A weak +1.

>   - ship a server.conf with:
>       no-auto-default=*

+1

>       ignore-carrier=*

A weak -1.

>   - tweak the ignore-carrier behavior so that if an ignore-carrier
>     device comes up, and it has no active connection, and a DHCP
>     connection could be activated on it, then we activate that
>     connection.

+1

> And for "future":
>
>   - further improve the ignore-carrier/DHCP behavior, so that if
>     an ignore-carrier device comes up, and it *does* have an active
>     DHCP connection, we renew the lease without taking the device
>     down.

It's a little bit against the meaning of 'ignore' but sounds reasonable.

Actually, I remember we talked about differentiating between the activation role of carrier and the 
deactivation one. What was the outcome? Is it reasonable to have a configuration that waits for carrier when 
activating but keeps the connection after losing carrier? Also, there's the Ethernet 4 second timeout which 
would be a good adept for a configuration option ranging from 0 to infinity.

Pavel