* Michael Stone said: > On Tue, Jan 25, 2000 at 06:03:10PM +0100, Marek Habersack wrote: > > > Every time I'm reminded of bind attaching itself to each interface > > > explicitly, I wonder why it does that (and doesn't simply bind to > > > 0.0.0.0). _Is_ there a valid reason? > > A router serving DNS for two LANs, for example. > > That's not an answer. OK, I'll elaborate on that. One copy of bind is configured to server requests (by default) on every interface attached to the machine. That can be useful in many cases. For example if you have a machine that routes packets between two networks (say, it's a firewall like that one described by the famous AT&T paper). There's no need to setup two separate DNS servers, instead you can have one copy listening on all interfaces. Another example is bind ran on a machine with dynamic interfaces (dial-up, PCMCIA, tunnels etc.) and it is desirable that all interfaces have bind attached - for example the machine is a gateway for the local LAN, but connects to the net using PPP. Then the bind is used to forward the local requests over the PPP link to the forwarder on the other side - it must attach to the newly created interface for that to work. The default setup guarantees that all these scenarios will work. It's a trivial task to change the default. marek
Attachment:
pgpnFT9v6fGNX.pgp
Description: PGP signature