[Dnsmasq-discuss] feature requests...

[Carlos Carvalho]

Paul Chambers (bod at bod.org) wrote on 31 March 2007 23:03:
 >One of the things I value about dnsmasq is that its simplicity,
 >performance and small footprint.

Me too. And it already has a lot of features in spite of being so
small. That's what makes it impressive.

 >While dnsmasq could be extended handle the complex situation you've
 >described, I worry that some of those desirable qualities would be
 >eroded in making it do so.

That's possible, I didn't look at how it could be implemented.

 >Have you already evaluateed and rejected alternative approaches?

The dhcp interface list could be done with firewall rules on the
server. However, at first glance the dhcp interface list seems to be
not too intrusive. Besides, the difference between dns and dhcp
behavior breaks the symmetry, and I like symmetry :-) I've even
already asked here why it's asymmetrical, and Simon said it's for
historical reasons.

The subnet-host file looks more invasive. Perhaps it could be
implemented with an extra level of indirection/hash in the table
lookup? The impact on performance would be small but not zero.

For my particular case this feature could be substituted by a flag
telling dnsmasq to only return dns entries for machines in the same
subnet (for that domain only, which is already accomplished by local
/domain/). It'd just be a more stringent filtering than what it
already does with localise-queries. I'd be happy with this but it
wouldn't be as flexible.