> * Lutz Hamann <l.hamann@diompika.org> [2007-07-28 23:24]:
> > I wonder if this is really the best solution, since not all drivers
> > support link_state and also tools like bgpd just use !LINK_STATE_DOWN
> > as indication that a link is up.
>
> yup, and you kinda gave the reason already, not all drivers handle link
> state correctly. some hardware is not even capable of reporting it.
> soin bgpd's case where we use the link state for nexthop verification,
> we check wether it is not down - which means it is either up or we
> don't know, in which case we cannot do anything.
>
> the problem for trunk is different tho, I am not sure what the best
> solution is.

it is clear that what works for one thing is not exactly what something
else wants.

but in any case, it is clear that some consumers (routing daemons)
need this per-interface flag to be less of a flappy bat, so that is
why the notion of "is up" was chosen instead of the notion of "not
down".

And I think for now it still is the right notion. Perhaps the trunk
code can analyze the interface status directly instead of using this
macro, plus other information, since the trunk code likely also wants
some level of "not a flappy bat". This macro, I believe, should not
change.