On Tue, Aug 5, 2008 at 1:40 PM, Simon Kelley <si...@thekelleys.org.uk> wrote:
> L Forrister wrote:
>
>>>
>> As far the leases file is concerned, doesn't the mac address (and/or the
>> client id) provide that differentiation?   I mean, having the same
>> unqualified host name on two separate leases would not cause identical
>> lines/records in the lease file.
>
> You're right, it wouldn't. removing the name from the leasefile is just
> an implementation detail: a way to implement the policy for DNS.
>>
>> The dns code, seems not to have a problem with multihomed hosts in
>> either /etc/hosts or /etc/dnsmasq.hosts (my dnsmasq addn-hosts file.)
>>
>> oso:~
>> lf$ dig +short -x 192.168.160.67 -x 192.168.160.68 -x 192.168.160.69
>> alkix.rebel.lan.
>> alkix.rebel.lan.
>> alkix.rebel.lan.
>>
>> oso:~
>> lf$ dig +short alkix.rebel.lan
>> 192.168.160.67
>> 192.168.160.68
>> 192.168.160.69
>>
>
> It works like it does because the one time when this actually happens to
> me is when I move my laptop from a wired network to a wireless one which
> is a different subnet. Under those circumstances I don't want DNS
> lookups for the laptop's name returning both the current wireless
> address, and the old wired one. Anything trying to connect would use a
> dead address for half the time until the DHCP lease on the wired network
> expired.

However useful sounding, I submit to you that this is broken.

If you are connected on wireless, you get a lease, the name is
associated to wireless.
When you connect on wired, the name is associated to the wired address.
When you disconnect from wired, the wireless lease may still be valid
(why wouldn't it be?), so the dnsmasq server isn't contacted and
doesn't update the name back to the wireless address.

>
>
> Cheers,
>
> Simon.

Reply via email to