> Yes it is (See RFC 2131 on DHCP protocol). DHCP doesn't require an administrator
> to manualy configure each new network device. That was the whole point of
> developing DHCP.
It's true that DHCP is intended to avoid manual configuration of
individual devices. *It* provides an IP address (and other
information), which means that the administrator doesn't need to
configure the device directly. It also, however, provides for address
"leases", so that the device doesn't necessarily get to keep using
that address indefinitely. The device is required to contact the
server to renew its lease occasionally, and must be prepared, if the
server so instructs, to stop using the address and start using a new
one. A reboot of the device is, thus, not required, and neither is
direct configuration.
[You can run a DHCP server such that it gives out infinite leases that
don't need to be renewed. However, this is functionally equivalent to
BOOTP.]
In short, DHCP is a way to configure devices over the network, and one
of its features *is* the ability to change the IP address on a client.
This does, indeed, cause active connections to fail.
-- Lowell Gilbert
> Eric.
> > > I assume here that you get a new IP after a system reboot since changing the
> > > IP while having an active connection is a serious violation of the DHCP
> > > protocol.
> > No. It's not. Changing IP addresses on clients without needing
> > to reboot them was the whole *point* of developing DHCP, as an
> > improvement to BOOTP.