Re: disabling keyring

[Antony J Mee <A J Mee ncl ac uk>]

Derek Frye wrote:

> I think the "no need to login" idea this is outside the scope of 
> NetworkManager; it's a *desktop* application, not a server application.

But that is silly.  I feel Eric needs more support here :-)
I have exactly the same feelings.  System and user based configurations 
are entirely valid and meaningful concepts whether a machine is some 
kind of dedicated server (eg. fixed IP machine) or on a desktop client 
machine (dynamic IP - ie. relevant to NM).

It is also absolutely reasonable for a client machine to have a server 
process listening on an port.  sshd is the perfect example! 

Perhaps some kind of nasty 'su'ing interface to have a list of 
/etc/init.d services which are started/restarted when the network comes 
alive, and optionally stopped/restarted when it disappears (some 
processes bind to localhost too)?  That would certainly be the 'client 
oriented' solution.  But they feel much more hacky than what Eric is 
suggesting.
And while I agree that some clients like ntpd should become dbus aware 
it does not seem quite so reasonable to ask critical service daemons 
like sshd to do so :- I suppose it could reasonably be an enable at 
compile time option for sshd, but we don't all run Gentoo :-)

Perhaps the correct extension to this idea is to have a completely 
independent dbus aware process that is capable of 
starting/stopping/restarting various system services
in response to various messages from NM...?

tOnY


> -- Derek
>
> Éric Brunet wrote:
>
> > On Wed, Oct 26, 2005 at 10:01:57AM -0400, Dan Williams wrote:
> >
> > > > Dropping into detailed mode is merely an aesthetic consideration, 
> > > > but presumably whatever ntpd is failing to do (update the system 
> > > > time?) is a technical issue, and I imagine there are other services 
> > > > that could benefit from having a network connection available to 
> > > > them once the network service starts up (fairly early in the 
> > > > service startup sequence).
> > >
> > >
> > > This is a hack.  These network services, like ntpd, need to deal with
> > > differing and/or absent network connections.  That's no different than
> > > now.  The only difference is that a network connection might not be
> > > immediately available to the daemon.  So when one _is_ available, the
> > > daemon should then try to contact the NTP server, but not before.
> >
> >
> >
> > Sure, I agree. ntpd, kweatherd, etc. are stupid processes that need 
> > to be
> > fixed.
> >
> > But ntpd is a client, network-wise. I was thinking of a server process,
> > such as httpd or sshd. If I run such a process on a computer, I 
> > expect to
> > be able to power up the computer, leave it alone a couple of minutes
> > (boot time) and be able to access the server from the network. It is not
> > so with NetworkManager: I need to power up the computer, leave it 
> > alone a
> > couple of minutes (boot time) and then log in to have the network 
> > working
> > and my server usefull. It doesn't matter whom I am logging as, just
> > someone needs to be logged in! It doesn't make any sense; why should
> > I need to log in as joe to have the system wide httpd working on that
> > computer ? And imagine that I just do that and leave the computer 
> > powered
> > up at home and go for a walk, knowing that some people are accessing the
> > httpd server. Now, if there is a power outage and the computer reboots,
> > noone will be able to access the server till I go home and log in again!
> >
> > That is a NetworkManager issue, completely specific to NetworkManager.
> >
> > Éric Brunet
> > _______________________________________________
> > NetworkManager-list mailing list
> > NetworkManager-list gnome org
> > http://mail.gnome.org/mailman/listinfo/networkmanager-list
>
> _______________________________________________
> NetworkManager-list mailing list
> NetworkManager-list gnome org
> http://mail.gnome.org/mailman/listinfo/networkmanager-list