On 2016-06-09 12:10, Rainer Weikusat wrote:

    [snip]

> The additional two lines disable the spanning-tree protocol. That's
> usually sensible unless one deal with a complicated, physicial topology
> which may or does contain loops. Then, it's needed so that the bridge
> can work out a loop-less virtual topology utilizing the existing
> physical one.
>
> Interfaces supposed to be used as bridge ports don't have to be up. The
> bridge manages this, IOW, it will initially disable all newly assigned
> ports and enable or disable them as required to implement the
> negotiated, virtual topology (if STP is being used) after the
> 'forwarding delay' (bridge_fd) has passed.

Well put.

And to assign an IP address to bridge:
- change "manual" to "static"
- add (at minimum) the address & netmask statements

like so:

auto bridge
     iface bridge inet static
     bridge_fd 0
     bridge_stp 0
     bridge_ports none
     address 10.0.0.1
     netmask 255.255.255.0

That also looks better than my previous example using pre-up/post-down
to addbr/delbr.

Rainer, since it wasn't in your previous example, I'm curious to know
your opinion on the use of "bridge_waitport 0" ?

Thanks.