[rbridge] IP data plane (Was: Use of 802.1ah Encaps)

[Gustavo J. A. M. Carneiro]

  Hello,

  Sorry to intrude, I just wanted to play devil's advocate here.. ;-)

On Ter, 2006-12-12 at 16:51 -0500, Gray, Eric wrote:
> Joe,
> 
> 	It's a huge drawback, and it requires configuration of
> at least IP addresses - possibly per-interface.  That is not
> going to work very well as a plug-and-play solution.

  An IP based data plane does not imply that you need configuration, not
even IP addresses on interfaces.  Think of IP adhoc protocols
(AODV/OLSR/etc.); they derive IP addresses from the L2 MAC address, or
simply generate a random IP.

> 	In addition, as you hint at, you're effectively now a
> part of the routing L3 network - as opposed to part of the 
> bridging L2 network.  It would make little sense - at that
> point - to try to differentiate RBridges from routers as it
> is already established that routers can be built and used 
> in real networks.

  There is still some advantage: normal routers need configuration,
RBridges don't.  But at the same time you can reuse existing IP
forwarding blocks of routers.

  But if you are really looking for arguments against IP data plane:
   1. Larger header size;
   2. Larger (32-bits) addresses (vs. 16-bits of rbridge nick names);
   3. Obfuscation, e.g. some fields are there but not used by RBridges
(like fragmentation control).

  Regards,

-- 
Gustavo J. A. M. Carneiro
<gjc at inescporto.pt> <gustavo at users.sourceforge.net>
The universe is always one step beyond logic.