Re: dnsmasq

[Howard Chu <hyc symas com>]

Dan Williams wrote:
> The idea is good and we should enable this sort of thing in the future,
> but we need better integration here.
>
> 1) NM needs to write out an /etc/resolv.conf that points to 127.0.0.1
> when people are using a local caching nameserver

I don't think the NM process should do this. It only makes sense if NM is the 
one that starts/stops the caching nameserver. Otherwise, it should leave 
/etc/resolv.conf alone in this case. (And I don't believe NM should be 
starting/stopping the cache server...)

> 2) What do we do when some other process (bind or a non-dbus-enabled
> dnsmasq) is already bound to port 53?  What's the failure mode here?
> Fall back to writing out a real resolv.conf?

Yes, there's really no other choice. I don't see that this is something NM 
needs to detect though.

> 3) Need some method of telling NM to use a local caching nameserver,
> either via the command line or via the system settings service or
> something like that

I guess an explicit switch/setting to enable this feature makes sense. Then it 
would be a fatal configuration error if the DBUS listener isn't present when 
NM starts up. Should NM abort its startup in that case?
--
  -- Howard Chu
  CTO, Symas Corp.           http://www.symas.com
  Director, Highland Sun     http://highlandsun.com/hyc/
  Chief Architect, OpenLDAP  http://www.openldap.org/project/