Darren Reed wrote: > If I'm running two different inetd's, one in a chroot'd directory and > one in /, > I want to have two different smtp services, each with a different > bind_addr.
So a Zone ? >>> What about adding "bind_interface" (at both the global inetd >>> level and on a per service basis) ? bind_interface is not a good idea IMO. Particularly if you consider things like IPMP and all the changes that Clearcase is bringing along. IP addresses are sensible, interfaces are a hardware concept and really shouldn't filter this high up into the stack IMO. >> Is this a common use case / request? Seems like a fairly unfortunate >> way to define configuration for inetd. > > Not really. In every case where someone uses an IP address, they > could also be using an interface name. I disagree IP addresses are not 1:1 with interface names. In addition to what I said above also consider Zones. > For example, I've often used "127.0.0.1/smtp" in inetd.conf and have > hacked inetd to support this. Why did I do this and not interface name? > Because it was easier to write the code to just parse the address and > put it into the sockaddr_in being passed off to bind than try and write > all of the code necessary to handle an interface name there instead. > > Had I the time and motivation, I would equally like more to be able > to use interface name as it removes associated problems created by > trying to provide services on DHCP controlled interfaces. Is DHCP the problem or is the problem actually that your IP address is not static ? This sounds like it is something that the NWAM project should be considering. -- Darren J Moffat
