Not to throw more coals on the fire, but another use for having multiple
IP addresses bound to the same ethernet card is for hosting virtualization.
Now I am going to ask a fairly dumb question -
In the below example, /dev/net/ipro1000/0 is a device, but it is bound
to an IP address. Who ever would want to read from or write to that
device? Is it sort of like the disk dev entries where they are (nearly)
never used by anyone but other, internal users (like the filesystem in
the case of disks or the network stack in the case of net entries)?
If so, what value is there in actually exporting them? Is it just to
have a common interface between drivers and the network stack and to be
consistent? Or is there some app/command line that will actually want to
ever access these?
Okay, so I'm proposing the following:
- if you create an interface /dev/net/ipro1000/0 it will be the first interface for this device
- when you want another interface (with a different address), you would need to create another interface called /dev/net/ipro1000/0:n where n is a number greater than 0 (since that will refer to the first interface).
- you can also refer to /dev/net/ipro1000/0 via /dev/net/ipro1000/0:0
IMO, that would make it simple and straight-forward to use. Is everyone okay with this?
Bye, Axel.
