Re: laugh test - capwap

["James Kempf" <kempf@docomolabs-usa.com>]

> What's more, if one portion of the "split AP" does in fact meet the
> definition of a router, I can't see how  both entities comprising the
Access
> Point can function correctly within the 802.11 architecture.   For
example:
>
> 1. What does a router do when it receives a packet?
> 2. If one half of the "split AP" did this, then forwarding decisions would
> most certainly not be made based on the learning table created from
> Association/Reassociation state, which is how an 802.11 AP functions.
>
> I think this highlights some of the risks in having the architecture work
> done in the IETF.
>

As a point of reference, an architecture for a router that supports a
CAPWAP-like architecture is described in the following paper:

    Zerfos, P., Zhong, G, Cheng, J., Luo, H., Lu, S., Li, JJ., "DIRAC: A
Software-based Wireless Router System",
    Proceedings of Mobicom 2003, pp. 230-244

The authors describe an architecture and implemenation, and give
measurements on three enhancements they implemented using the distributed
routing function: FMIPv6 for fast handoff, using FEC instead of retransmits
to eliminate head of line blocking, and policing. The DIRAC architecture
does not, however, include tunneling nor security.

            jak