new CARP implementation

[Gleb Smirnoff]

On Wed, Aug 10, 2011 at 09:38:04AM -0700, Freddie Cash wrote:
F> However, I'm not sure I understand the reasoning for removing the carpX
F> pseudo-interface.  It's really nice having the symmetry between carpX,
F> vlanX, brX, and other pseudo-interfaces, and keeping the configuration
F> details separate from the underlying physical interface.
F> 
F> This now makes creating/configuring CARP different from creating/configuring
F> VLANs.  :(

This is done because VLANs _are_ interfaces, they are tunnels within ethernet
interfaces, splitting one interface into a bunch. Bridges are interfaces, as
well as LACP trunks (lagg(4)), since they group a number of interfaces into
one. CARP addresses _are not_ interfaces, they are addresses. IMHO, implementing
them as virtual interface subtly attached to a real one, was a layering
violation.

-- 
Totus tuus, Glebius.