[rbridge] Time to summarize "forward or block" BPDU thread

[Stephen Suryaputra]

Joe,

To me, Rbridge is changing the assumption that there is no router in
the L2 network. It fits well with the way L2 network view of routers,
they are nodes. From a directly connected L2 network, the Rbridge is
sourcing/sinking packets, and blocking (but accepting) broadcast. But
instead of terminating them (like nodes), Rbridge is tunneling them.

Another point why I think they are nodes, tunneling (i.e. the shim) is
using the unicast address of the ingress and egress Rbridges in the L2
header (no?).

Since the goal is to reduce the size of spanning tree protocol
network, the Rbridge chose not to tunnel BPDU. Tunneling is an option,
but that will make a cloud of Rbridges to be a virtual single bridge,
which IMHO can create more complex control for Rbridges. Tunnelling
BPDU however makes Rbridges look link links, very clean
architecturally.

So, now there are 4 devices in the L2 network: nodes, links, bridges
and Rbridges.

Cheers,

Stephen.

On 10/17/05, Joe Touch <touch at isi.edu> wrote:
>
>
> Radia.Perlman at sun.com wrote:
> > It works perfectly well if RBridges block. It isolates the bridge islands.
> >
> > Routers also block bridge spanning tree, and to exactly the same
> > effect as if RBridges block BPDUs.
>
> This isn't the same at all; I keep hearing the analogy, but it fails
> under all tests:
>
>         first, there are NO routers on an L2 net
>
>         all sources/sinks are the same (called nodes AFAICT)
>         and hosts and routers are just nodes,
>         identical as far as L2 is concerned
>
>         nodes block all BPDUs (we agree), but
>         nodes also block broadcast (rbridges don't)
>
> Near as I can tell, L2 has exactly three devices currently:
>
>         nodes
>
>         links
>
>         bridges (participate in SPT, interpret BPDUs)
>
> Bridges don't 'block' BPDUs, but they don't forward them to all ports
> either.
>
> Now, an rbridge can be "like" a node, a link, or a bridge and still play
> in the existing L2 specs. If it isn't 'like' any of those, then we're
> changing the definition of 802, and I didn't think that was the goal here.
>
> rbridges - at least to nodes on the L2 - are NOT 'like' hosts, since
> they don't source or sink L2 traffic to existing nodes
>
> rbridges could be 'like' links if they are TRANSPARENT
>
> rbridges could be 'like' bridges if the PARTICIPATE
>
> What exactly are rbridges 'like' if they BLOCK?
>
> They can't be 'like' routers - there are no routers in L2.
>
> ??
>
> Joe
>
>
> _______________________________________________
> rbridge mailing list
> rbridge at postel.org
> http://www.postel.org/mailman/listinfo/rbridge
>
>
>
>