host statement scope rules (ISC DHCP 3.0.5b1)

[Ted Lemon]

Glenn Satchell wrote:
> Based on the number of posts to this list it appears that unfortunately
> it is common practice to configure hosts entries this way. Enforcing it
> in the way you propose would break a lot of configurations. Printing a
> warning is a friendlier way to do this.

Furthermore, if you know what you're doing, expressing a host statement
in a non-global scope isn't an error.   You need to remember that in the
DHCP configuration there are actually two sorts of scope: IP address
configuration scope, and IP option configuration scope.   The two are
tied together, but they aren't the same.

The mistake people usually make with host declarations is to think "If I
declare this in a subnet context, then it will only be in effect for
hosts connected to that subnet, and will be ignored otherwise."   This
is not true: a host declaration matches the intended host if no fixed
address is present, or if a fixed address that is present is valid on
the network to which the host is connected.   And for no other reason.

However, a host declaration in a non-global context inherits the options
and settings of the context in which it is declared.   So moving it to
the global scope is not a neutral act - it's going to inherit different
configuration settings and options if you do that.

It might be better in the abstract to get rid of this dichotomy, but at
this point I think it would cause as much confusion to get rid of it as
it does to leave it the way it is.