[rbridge] Updated problem statement
Erik.Nordmark at sun.com
Fri Jul 9 08:30:06 PDT 2004
Based on some editorial comments from Hilarie Orman I've tweaked the
wording in the problem statement. Hopefully it is more succinct as
Organizations with large local networks may wish to administer
them without the complications of internal IP address assignments.
Layer 2 addresses provide the unique names, but the most common
layer 2 interconnection method (dynamically created spanning tree
formation using bridges) is not as flexible as layer 3 routing.
The BoF will explore hybrid solutions that retain the simplicity of
administration while taking full advantage of complex topologies.
Bridges with spanning tree algorithms have obvious disadvantages: routing
is confined to a spanning tree (precluding pair-wise shortest paths),
ARP and Neighbor Discovery packets must be carried across all the links,
the header on which the spanning tree forwards has no hop count,
spanning tree forwarding in the presence of temporary loops spawns
exponential copies of packets, nodes can have only a single point of
attachment, the spanning tree, in order to avoid temporary loops,
is slow to start forwarding on new ports, and it is not possible to take
advantage of the rich physical topology for capacity since the packet flows
are restricted to following the spanning tree.
Routers avoid those disadvantages but have their own problems:
IP addresses are link specific so a host that moves must change
its IP address, the routers must be configured with unique link prefixes
for each of the attached links, and the block of IP address space can not be
fully utilized because it must be partitioned across the different links.
The BoF will explore combining benefits of bridges and routers in a way
that will co-exist with existing hosts, IP routers and bridges. The
design should support both IPv4 and IPv6.
More information about the rbridge